BIOMECHANICS II BMED 4962
Popular in Course
Popular in Biomedical Engineering
This 228 page Class Notes was uploaded by Joshuah Labadie on Monday October 19, 2015. The Class Notes belongs to BMED 4962 at Rensselaer Polytechnic Institute taught by Staff in Fall. Since its upload, it has received 34 views. For similar materials see /class/224825/bmed-4962-rensselaer-polytechnic-institute in Biomedical Engineering at Rensselaer Polytechnic Institute.
Reviews for BIOMECHANICS II
Report this Material
What is Karma?
Karma is the currency of StudySoup.
Date Created: 10/19/15
BMED4962ECSE4962 Introduction to Subsurface Imaging Systems Welcome and Introduction Kai E Thomenius1 amp Badri Roysam2 1Chief Technologist Imaging Technologies General Electric Global Research Center ZProfessor Rensselaer Polytechnic Institute GE Global Research Center for SubSurface Imaging amp Sensing Basic Class Data BMED496201 ECSE496201 CRN 54506 3 credits Instructors Dr Kai Thomenius Dr Badri Roysam Tuesdays amp Fridays 1230 150 Classroom JEC 4104 What this course is about You can expect to learn the basics of imaging objects that are hidden under a surface in the context of realworld applications Along the way we ll also learn about exciting applications requiring subsurface sensingimaging Medical Imaging Systems Industrial Inspection Systems Bio imaging biology biotechnology Motivation for Medical Imaging Healthcare costs Clinical ef cacy Enormous spend 18T in US 50 ofheart failure death 265kyr 39 Adm 505 530 325 USquot healthcare spend on strokes Adverse drug events 770K us deaths System cannot handle excess capacity or injuries years Shrinking population of doctors I 39 0 Chmn39c cm 70 f st 9 people100 unintended infection Huge markets 35003 H 1005 305 magnum Dingnnshcs Haathan 3721123mmmmmm mmw We a t s Trudmunul amude Phunwu elobui Vmugimg appammy Spend w m N Healthcare Trends Drive Imaging Growth e4 pup 65 Aging Pupulntion Globally mu m Erm leslm am 5 r nisease iiiaiaeiiae mairsi Disease Trend iraiii shapes to Molecuhr peaaie aver uie age of as require 3 a z iiiries gs iiieiiy imaging investigations as mass under the a eauis Cancer Heart disease 1 iiiiiers e Cardiu13 ofull aemris china 7 v2 Europe 7 Tap cause 40 Us Obesity on the KISS More pntlems wiui Miiiupie Diseases Advent of dlugnus crtll iven disease management Technulugy rasigii shows anatomy moieeiiiar mergpeiicie ef wa Improved image gaaiiiy for better diagnoses Faster exam nines sriima outpatient and noanVOSIVE HOW can we look under surfaces Computer mage Decision Features Coordinates other Surface 1 Probes Detectors iiiei i wm Takes ideas from multiple disciplines working together What subsurface problems are we talking about 0 Imaging MMCTUltrasoundOptical Sensing Decision making based on acquired data Characterization Tissue elasticity Spatial registration Procedure guidance eg biopsy mine location Examples of Imaging CT Small Animal Imaging Coronary Artery Visualization w MRI Ulcerated Plaque Cadaver Heart w CT Examples of Sensing USMammography Fusion Luggage Inspection Example of Positioning Jkgistration rush Ultrasound Applicator Treatment Plan Ablated Tissue l 40 mm Example from Noninvasive Nonincisional Surgery Determination of Features of Target under Study Comparison of co registered sonoelastography and b mode ultrasound images with a nearby histology slice 01 12 case 015 11683 5 04 Tissue Elastic Properties Interaction between the probe and target conveys the information Terahertz Scattering Homeland Security Types of Probes For the purpose of this course the term Probes describes the means eg radiation used to acquire imaging amp sensing data Hmoherent i uerent C assim wtside Auxi ierry How Probes Interact with Media um meuium Multiple Scanering Semen rg Tur bulent Medium Multiple Propagation diffraction When can we see the subsurface object Computer Image Decision Features Coordinates other Surface 1 Probes Detectors me A s 1 sf liem ulm 1 Probe should reach the object 2 Enough signal should reach the detector 3 The detected signal should be affected by the object Example Example Xrays go readHy through tissue but are attenuated by K Absorption Absorption Dispersion W Molecular Phenomena Sometimes I I Optical 3D393C0 f03ra l Microscope TUmoriBla od Vegsels quotw w Ulcerated Plaque Cadaver Heart w CT Coronary Artery Visualization w MRI Common Characteristic to Probes Wave Physics Subcellular Biology Tissues amp Organs CG 100nm 10 m 100 pm 10cm Undergroqnd m J 3r Underwater DlagnOSlS 5L5 Wm Exploration Outline of Course Topics THE BIG PICTURE PULSE ECHO METHODS What is subsurface sensing amp Examples imaging MRI Why a Course on this topIC A different sensing modality EXAMPLE THROUGH from the others TRANSMISSION SENSING Basics ofMRl XRay Imaging MOLECULAR IMAGING Computer Tomography What is it COMMON FUNDAMENTALS PET amp Radionuclide Imaging Propagation 0f W3Ve IMAGE PROCESSING amp CAD interaction of waves with targets of interest What We Expect From Students Prerequisites ECSE2100 Fields amp Waves 1 ECSE24lO Signals and Systems Willingness to learn some Biology Medical Terms Physics Math Computing Prior exposure to computers Internet and basics of programming Course content reasonably exible Expectations assignments etc proportional to prior background Choice of topics and coverage to meet student interests Grading Assignments 60 To be handed in within one week Project 40 teams up to 2 ok Indepth review of a speci c course topic e g processing of raw data sets Topic subject to instructor approval All students should show independent homework and projects No exams Generous reward for effort evidence should be shown Study Materials Lecture notes Other class handouts Selected internet sites MATLAB We have campus licenses at Rensselaer UPRM NEU amp BU Field Trips Several eld trips are being planned RPI Biotech Center GE Global Research Albany Medical Center Others Goal Direct observation of subsurface imaging Summary Subsurface imaging Course Focus Many applications to Biology Medicine Industry amp Homeland Security Probes media and probemedia interactions 0 Methods by which we examine contents of hidden volumes Wave theory the common thread 0 Common to electromagnetic and acoustic probing Next Class Overview Projection Imaging CenSSIS Maj or Research Center on Campus Mission Create a uni ed engineering discipline for sensing and imaging objects that are hidden under surfaces Diverse Problems Similar Solutions B enefits Learn from similarities 0 It s all the same physics Learn from differences wwwecserpieducenssis Why is an Xray beam different from ultrasound Learn from experience with other probes GE Global Research Niskayuna I Advanced Research for all of GE Founded in 1900 in a barn in Schenectady by Steinmetz Imaging Products amp Services GE Medical Systems 0 CT Functional Imaging MRI Ultrasound X Ray Support PACS Workstations GE URLs wwwgecom wwwgehealthcarecom Instructor Contact Information Badri Roysam Professor of Electrical Computer amp Systems Engineering Of ce JEC 7010 Rensselaer Polytechnic Institute 110 8th Street Troy New York 12180 Phone 518 2768067 Fax 518 27662612433 Email roysamecserpiedu Website ht www ieduro sab NetMeeting ID for offcampus students 1281136180 Secretary Laraine Michaelides JEC 7012 518 276 8525 mm quot Rensselaer why nal change the world Instructor Contact Information Kai E Thomenius Chief Technologist Ultrasound amp Biomedical Of ce KWC300A GE Global Research Imaging Technologies Niskayuna New York 12309 Phone 518 3877233 Fax 518 3876170 Email thomeniucrdgecom thomeniusecserpiedu Secretary Laraine Michaelides JEC 7012 518 276 8525 mm GE Global Research quot 39 m M Mg m mm BM ED 4962ECSE 4962 Introduction to Subsurface Sensing and Imaging Systems Lecture 23 3D amp MultiSpectral Microscopy Kai Thomenius1 amp Badri Roysam2 1Chief Technologist Imaging Technologies General Electric Global Research Center 2Professor Rensselaer Polytechnic Institute GE Global Research Center for Subsurface lrnaging amp Sensing 5m 1 35 Recap of Last Lecture 3D Scanning Microscopy The multiphoton effectis a basis for 3D imaging Second Harmonic Generation Imaging SHG is a lossless kind ofmultiphoton microscopy Fluorescent Proteins and multiphoton is a magical combination that allows livecell Imaging We can do 3D microscopy without the multiphoton effect Need to understand the optical properties oflenses Magni cation Resolution axial and lateral Next Class Confocal Microscopy MultiSpectral Imaging for uorescence multiplexing Reference quotMicroscope basics and beyondquot by Mortimer Abramowitz i w inn i ii quotOptical Microscopyquot by Michael W Davidson and Mortimer Abramowitz Website quotMolecular Expressionsquot sud 2 I35 Recap Lens Properties Angular Axial inreusiry Distribution Aperture NAnsinu 1 MM NA 2 Numerical Aperture 1 051x 2 4 11 is the refractive index N39A39 Lateral resolution Axial resolution lieu35 Obtaining 3 D Structure without the Multi Photon Effect quotConfocol Microscopyquot Nome comes from quotconjugate fociquot of lenses f1 and f2 Bosic ideo Put 0 tiny pinhole ot the conjugate focus f2 Almost all of the light from the point in the specimen quotsqueezes throughquot the pinhole Pinhole f2 1 Pornt 1n the specrmen Shdz4l35 Light from Point quotBquot 0 Basic idea cont 7 A most GM of the tht from the point quotAquot in the specimen quotsqueezes through the pinho e 7 Most of the tht from point quotBquot is b ocked by the pinho e aperture 7 SimHorstory fora point betow A mm The Confocal Microscope mmting 39 L I39CEH llll min39orx hm pmllulu v dulcclm I M39l t mamMme lmmccm mmple sham35 The Confocol Microscope Photomultiplier Detector Detector Pinhole Aperture OutofFocus Fluorescence Light Rays Barrier Finer Excita on Filter lnFocus Ugh Rays Laser Exclralion Source Excitation ngm I 1 Rays Ll ht Source inhale Aperture Focal B Planes A Specimen shag 7 Us Effect of Pinhole Mam Imnna cquot39E mama Image Cnmnczl Image 17 Fncusmk Chnnsansrecmm Magnmcmmn 1 I7 Fmsmk ChnnseASreclmen Maunmmn P nllen Gern jv Mmmm 55 55 Fucus BFIHMHESS ZJ XIS P S l quot PMY REquot 6339quot Fncus Brluhlness Zrnxls Pnsmnn PM Red 62m Medium 33 M u 33 Plnhnle Auenure 512 Cw thnle Auenure Size F Small F Medium 0 Lzr e quot r n Scan Llne Speed PMI Green 62m small MEmum mg Scan Line Speed pm Green 62m shag 2 Us Effect of Pinhole Wldn ald Imm comai ima ChoosoA Emma L i 1 Focus my Mngnmeauan LeafAbs sslm Facus mm mum mumquot Medium 2539 PlnhaleApemsreSke 5mm MEE UN a La Scanune Speed FMYGreenGaln 1m iiwwwmmm mn39snedu xmexlwmAlcam acalmdethl m y 35 Opticol Slices Serial Op cal Sec nns by Donmcal Microscopy Move thezstage Lipdown At each 2 value scan across the gtltey eld to collect an optical slice A stacllt ofoptical slices is a volumetric BeDl image 51mm How do we view a 3D image 0 The computer screen is two dimensional so we have to project the 3 D volume Igtltyz onto a 2 D image Pgtlty Two basic choices to make 0 Projection angle 7 Simpiest Aiigned Witn axes 7 Best Objiaue angjes 0 Projection formulaalgorithm 7 SumAverage Median 7 Max Min 7 Surface Rendering 106 y 2 m nzs Viewing by Projections The maximum value projection is most commonly used for uorescence data Neuron TR054Z1 Step Size 05 um Zoom 10 Dimension 512x480x323 Volumetric Rendering Chooseuvfewm an e Ray casting Shootfuys buck mto mevofumeffomeuch pfxef m PM 7 Cumpute an mtensmyvume by summuuufvmuxmm etc Good objectivesmcewe39fenot messmg Wf quot the dam a mputuuontuexpenswe 012 Co VE u en desffe 1 mm mega fmemmvefy m uukffum m efenmngfes nuumbesfuwunufegufuf umputEf e Expensfve m d euccefemmrs and pursue campumuun ufgunmms uvmmbfe fur reumme vu umefendefmg Practical Volumetric Rendering f r w in mf at h fffsmnd nu surfaces m the sen Wage e WE fEmEssmg memeccm Use gruphms mutmesm creme um cmf hgmmg effems m render the surfaces 7 VEryfustsmcE mus1 cumputErs hqusuppurtfur thfstypE uf gfcpm e fuccysmmegcme Eurds uutp2rfurm 2xp2nsw21UUK rEndEHng Enngs ffum l yEurs ugu 7 Pretty butputEnUtu msfecmng fmesurfcces cremenmec muncufumy Need m pay manus m unequuf uxfm utem resumuun Effective PointSpread of the Confocal Microscope 86 0 of the PSF overall x y 2 PSF illuminati on x 3 Z X Psalm04x y Z HJ Overall PSF is tighter than the PSF ofo lens ampquotcomet shapedquot Affeth lay Pinhole Size 1535 Resolution of the Confocal Microscope Ideal case When pinhole is Axial and Lateral Faint Spread Functions super tiny Le N 0 and the excitation and emission 1quot 39 xquot wavelengths are close g 5 39 xi PSFillumination x 3 Z E g Lat nl PSF N Es AxialPsF N PSEietec onx5 3 Z In Flame 5 5 39 g 5 11 The overall PSF is roughly V0pll n unlls the square of the lens PSF in this ideal case m m Resolution of the Confocol Microscope Conventional Microscope Confocal Microscope 13977X1Emission X R resolution 2 NA2 2 NAZ Lateral it resolution ny 051xw 12W 037x SM 17 l 35 Imoging Noise 0 Loser Noise Fluctuotions in loser power Not criticol for fluorescence mode imoging 0 Shot Noise Poisson Noise Due to the poucity of photons especiolly when the pinhole is norrovv Use longer sensing time or overoge of multiple imoges o Detector quotDork Noisequot Folse signol generoted by the detector in the obsence of photons SM 12 l 35 Deconvolution The Softwore Method to Opticol Slicing Observed Image True Image yxyz 139 Jay 2 Yuv w Iuv wHuv w Deconvolution Exomple softwore method Before Confocal Deconvolution Example Rat CAj Hippucampal neurun image Data Cuunesy Dr James Turner Wadswunh Center Albany Ng g kgs Comments on Deconvolution We see from the neuron examples that the software deconvolution method has the greatest impact an axial resolution the lateral resolution is not improved much It is great for imaging filamentous and tiny objects The software method is computationally intensive lots of 3D Fourier Transforms So when axial resolution is critical the software method is valuable amp worth the computation Slide 22 35 MultiSpectral Imaging 0 Basic Motivation 7 can We capture multiplefluorescent signals at once Fluorescence multiplexing We want to capture the relative SFOthl context oftwo or morefluorescenty a ee structures 0 Solution 7 Build instruments that allow us to adjust two things in unison the excitation wavelen ths 7 Put a lter Wheel in front ofa broadband source to select illumination Wavelengths 7 Use multiple andor tunable light sources the wavelengths that our detector is sensitive to spectrally resolved detection Array of detectors 7 Use optical lter Wheels in front araetectars 7 Use a prism or a airfractian gratin to split the detected beam and an array ariig t detectors D1 tion W 4 Grangera K The Zeiss A l 39 Detector META System An ay The fluorescence light after the pinhole is passed through a grating to an array of 32 detectors lPMT39sl Produces a quotlambda stackquot llx y 2 7t A spectrum at each pixel main 5Label Immunohistochemistry Nuciei Emission spectra LLK Emission spectra LPMk Dealing with Overlapping Spectra mum 3amp1 a 500 5m 5m 50 m sea 52 5390 5a mu 0314 mi arm am Nucleus histone GFP Fusion Actin laments uorescein conjugated phalloidin The peaks are separated by only 7nm sud 27 35 Dealing with Overlapping Spectra Reference Spectra 5w 5m mi 530 540 Vimu Unmixing Resultquot A and A am rear SUM21 141065 ZXR1l142xy ZXR2l Compute A1 and A2 at each pixel subject to constraint A1 142 1 sud 22 35 Confocal vs Multi Photon o Confocai microscopy is cheaper Causes rnare phatadarnage Causes rnare phatabieaching throughput the cane rather thaniust the neck Suffers frarn chromatic aberration different caiars carne to focus at different depths Faster Confocals The basic idea is to nmlens Spre aeda39l39aser scan more than one my Dinnmmau paint t Mlmr I Llne scannlng Plnnnle A 53 39 Scan an entire raw uf Array 7 I Lens I the WEIgs utunce Nlpkow DISkS Scans iuts afpaints at unce using a mmung disk win a spimi 39 Ohealve m i i i D I Ital LI ht Swnches Lsgecimen 9 9 Arrays afmirrurs Excirauon imensityiti Advanced Topics we39re leaving out Decay Flunxescenne uorescence Lifetime imaging WM 7 Aiso expioits the temporai response ofa uorescent rnoiecuie in addition to its spectrum Requires fastdetectors and eiectronics forreairtirne photon counting and processing uorescence Resonance Energy RET Transfer F 7 One rnoiecuie absorbs iight and transfers energy to a ciose neighbor1 e10nrni rnoiecuie noternie iig t 7 Measures rnoiecuie cioseness beyond Rayieigh iirnit Ultimate Optical Microscope of Space x y z e isotropic sampiing No wasted photohs the Future Compiete spectrum at each pixei e Meas ur e Compiete fiegtlti e Compiete fiegtlti emission spec Time respohse a e Photon counting mission spectrum biiity to shape the excitation spectrum biiity to capture and anaiyze the trum igtlt each p ei hardware at each detector 7 Time response at different spectrai waveiengths Muitipie modaiities oh ohe piat orm Homework 0 Slide 3 reviews the definition ofa Numerical Aperture a measure of the lateral focusing strength of a lens How does this measure compare to the f number we used with acousticfocusing What is their algebraic relation Slide 19 discusses the use of deconvolution for improving the spatial resolution of optical imaging systems How would you avoid the problem of having point spread functions with zeros in the regions of interest SM 33 l 35 Summary 0 We can do 3D microscopy without the multiphoton effect Confocal Microscope Computational Deconvolution Based Microscopy Visualizing 3 D images Multi Spectral Microscopy 0 Next Class Image Analysis Fundamentals Image Segmentation SM 34 35 Instructor Contact Information Badri Roysam Professor of EIectricaI Computer amp Systems Engineering Office JEC 701 Rensseiaer Poiytecnnic Institute 110 8m Street Troy New York 12180 067 Email roysam ecsergiedu Website nttg zwww ecse rgi eduZNroysam Course website nttg www ecse rgi eduNroysamCTIA Secretary Loraine Michaeiides JEC 7012 518 276 78525 micnai rgi edu GraderVing Cheri cheny9 Qrgiedu Office JEC 6308 51827678207 why nnl Ehl gl m nunquot 525 Center for 39 K ensmg Instructor Contact Information Kai E Thomenius Chief Technoiogist UItrasound amp Biomedicai Office KW7C300A GE Giobai Research 7 Fax 518 38776170 7 Email thomeniu crd ge com tnomeniusecsergiedu 1 i39quot Secretary Loraine Michaeiides JEC 7012 518 276 78525 micnairgi edu i Rensselaer GE Global Research m v mas BMED4962lECSE4962 Introduction to Subsurface Imaging Systems Lecture 5 Introduction to CT Scanners Kai E Thomenius1 amp Badri Roysam2 1Chief Technologist Imaging Technologies General Electric Global Research Center ZProfessor Rensselaer Polytechnic Institute GE Global Research Center for SubSurface Imaging amp Sensing Outline of Course Topics 0 THE BIG PICTURE PULSE ECHO METHODS What is subsurface sensing amp Examples Imaging MRI Why a course on this topic EXAMPLE THROUGH aglggreernst sensnng modality from TRANSMISSION SENSING XRay Imaging Basncs of MRI Computer Tomography COMMON FUNDAMENTALS What is it propagation of waves PET amp Radionuclide Imaging interaction of waves with targets o IMAGE PROCESSING amp CAD of interest Recap Three aspects to Xray imaging xray sources xray interactions with matter detectors analog digital Performance metrics MTF SNR CNR DQE forxray performance were given Justi cation for digital detectors was based on these ROC Gunes for hypothesis testing TP FP TN FN Modulation Transfer Function Contrast M T F 7 Contrastm Contrasttonoise ratio CNR 2391112 2391112 a W Detective quantum ef ciency sig 1 S I I DQEifJ39 A N YPSi HX 39 C39 Noise power Exposure spectrum mR Incident energy densitymRmm2 LambertBeer Equation 2 0 eH Z where z is the xray intensity at the measurement plane O is the xray intensity at the 1 source plane 0 v 2 gt z is the distance between the source amp measurement planes u is the attenuation coefficient Also known as BeerLambert or Beer s Law o The goal of the CT scanner is In general H 15 a to apply the LambertBeer function of Z this across the cross section XRay Imaging to CT Imaging i as L X ray limitations 3D structures are collapsed into 2D images Low softtissue contrast great for bones Not very quantitative X ray CT Similar hardware to ordinary X ray Image ofa slice extendable to 3D But heavy computational load Brief History of CT Scanners 1972 CT scanning invented by Godfrey Houns eld a UK scientist Announced computed axial transverse scanning Initially known as CAT scanners this caused predictable cartoons to show up Presented initial crosssectional images of the brain etc The invention was shown to have excellent diagnostic potential immediately Houns eld shared the 1979 Nobel Prize with Alan Cormack who derived alternative CT reconstruction algorithms Unfortunately Dr Houns eld s company EMI failed to capitalize on the invention and has since then left medical imaging Dr Houns eld passed away in 2004 I a AP M From CafePresscom Hounsfield s Apparatus CT Scanners This is what they look like today Basically rotating xray tubes amp detectors V th a lot of computing power CT scanners are being used for All types of medical diagnostics Airport luggage inspectors Nondestructive evaluation of materials CT System Components Types of Images Viewing Modes Coronal Rendering Sagittal I Courtesy of Tom Toth GEHC 5 oEmEogt wmm Image Reconstruction Problem CT images generated by a reconstruction from projections Projections can be understood on the basis of LambertBeer How can we generate an image given a set of such projections The image shows a single measurement of attenuation through a brain section In CT scanning many parallel measurements are made to form a projection ofthe attenuation Let us define our measurement g as I g 2 X ray tube CT Reconstruction Problem Let us make a model of the unknown object n boxes of same size but different attenuation coefficients Now applying LambertBeer we get 16 0 Z im 2106 i1 6 1 le zAx 3AX e e uz 2 3 4 5 gtAX I IunAx u CT Reconstruction Problem Using our definition 9 we have I n g 1470 ZyiAx i1 And in the limit as Ax goes to zero g Tyltxgtdx I p0 1 2 3 4 5 In y gtAX I CT Reconstruction Problem Our intuition tells us that a solution is possible Here s a quick demo to show that this indeed is the case Let us do four experiments in which we transmit four xray beams of intensity 0 and receive four times with intensities l1 l2 l3 and I4 We can express the results mathematically with the LambertBeer relation Thus we have 4 equations 4 unknowns problem solved CT Reconstruction Problem Unfortunately the real world is not as kind as the example implies For clinical utility we need at least a 512 by 512 pixel grid Thus we would need at least 264144 equations to solve for that many unknowns This is challenging to say the least when CT scanners were first introduced in the late 1970s this approach was used Images had to be calculated with fewer pixels people gave up on resolution so as to achieve reasonable reconstruction times 10 a 10 Projections amp Sinograms Sinogram Projection all rays in Sinogram 2D plot Of directiOn e are all projections as a summed along the rays funCtiOH 0f 9 and projection Width A er hnp39 dnlnhin radinln ninwa 39 TN 1 indp hcm CT Image amp Its Sinogram Basics 2D Fourier Transform 2D FFT of an image fxy A good way to understand CT reconstruction Actually whole bunch of 1 D FFTs In the image shown most energy in low frequencies why Assuming image size of 8 what is the frequency increment in FT V th N pixels what is the largest frequency in FT Transformed image is in k spacequot 0 Spatial frequencies 27 27 kx k x y Anew u 1 Fourier Slice Theorem proj ection filten39ng 1D DFT line of the 2D nal image Data am rt A er11u L A M L A I n OROF I r Fourier Slice Theorem I 7 XIV P0t Fuv Fourier Slice Theorem can be used to derive a superior reconstruction approach Here s what it is The 1D FT of a projection is equal to a radial slice of the 2D FT of the image With enough 1D FT s of the projections one can estimate the 2D FT of the image amp by taking the inverse the the nal image It also forms the basis for the filtered backprojection algorithm A er httn39 dnlnhin mdir n ninwz quot TPh lindP htm H 2 14me ng W 7 2 73 1jluijji 2 j yulm Backprojection with a Point Target Backprojection Alternative to Fourier Slice Recon Based on symmetry relations of 2D FT Once projection data are acquired one can begin the These steps are repeated for all projected angles Introduction of Filtering Backproj ection reconstruction w no ltering Impact of lter on Sinogram Backprojection reconstruction w lter compare images Backprojection Shepp amp Logan Phantom Backprojection Torso IIMI i llquot Summary Introduction to CT scanners Emphasis on reconstruction algorithms Basic problem of reconstruction from projections Simple reconstruction using linear equations Fourier Slice Theorem Backprojection amp filtered backprojection References ttgthayerdartmouthedubpoqueENGG167122520ROC25 20Analysisgdfampe7620 httpdolphinradioloqvuiowaeduqeSlidesCTPhys1indexhtm Homework 3 Algebraic Reconstruction Technique Determine the attenuations in the four pixels as follows With no prior knowledge assume uniform attenuation ie I1 IZ 4 for each of the four pixels Reevaluate horizontal line integrals compare new values w measured results eg u1u2 3 Split the error evenly w M and uz for top row u3 and M for bottom row Repeat the process for the vertical line integrals Repeat until no further improvements Try for different values preferably double digit 10 p 3 10 gt 7 4 6 ART was the first recon algorithm used may be making a comeback Instructor Contact Information Badri Roysam Professor of Electrical Computer amp Systems Engineering Office JEC 7010 Rensselaer Polytechnic Institute 110 8th Street Troy New York 12180 Phone 518 2768067 Fax 518 27662612433 Email roysamecser9iedu Website htt wwwr ieduro sab Secretary Laraine Michaelides JEC 7012 518 276 8525 michal r iedu Rensselaer why nut clung in warmquot Instructor Contact Information Kai E Thomenius Chief Technologist Ultrasound amp Biomedical Office KVVCsOOA GE Global Research Imaging Technologies Niskayuna New York 12309 Phone 518 3877233 Fax 518 3876170 Email thomeniucrdgecom thomeniusecsergiedu Secretary Laraine Michaelides JEC 7012 518 276 8525 mm Rensselaer GE Global Research m u mquot n mum BMED 4962ECSE4962 Introduction to Subsurface Imaging Systems Lecture 4 X ray Imaging Kai E Thomenius1 amp Badri Roysam2 1Chief Technologist Imaging Technologies General Electric Global Research Center ZProfessor Rensselaer Polytechnic Institute GE Global Research Center for Subsurface imaging amp Sensing Slide 1 Review of Last Lecture Methods for classifying different imaging probes 7 Coherence 7 Interaction with target 7 Projection or pulseecho Xrays as an example for those classi ers Today 7 Projection Imaging 7 Xrays Slide 2 Outline of Course Topics PULSE ECHO METHODS THE BIG PICTURE What is subsurface imaging Examples Why a course on this topic MRI 39 EXAMPLE PrDIECIIDn Imaging Adifferent sensing modality XRay Imaging 39om the othe s Computer Tomography Basics ofMRI COMMON FUNDAMENTALS MOLECULAR IMAGING What is it Propagation ofwaves Inter ct39on ofwaves with t a I PET amp Radionuclide Imaging targets of interes IMAGE PROCESSING amp CAD Slide 3 History Nov 1885 Rontgen discovers X rays in Wurzburg German 7 Observations of uorescence when using a Crookes tube Initial results published in Dec 28 1885 broader announcements in January 1 86 In the next several months reports of diagnoses of fractures bullets even uoroscopy started coming Late in 1886 Siemens amp GE started selling Xray equipment Flinn inn PELYIX Xray Generation Bremsstrahlung Electromagnetic radiation photons Wavelengths 7i ranges from 10 Low Energy A High Energy E pm to 10 m Low Frequency High Frequency Propagation speed cn 300000 krnsec Energy of xray photon 1w where h Planck s constant or e 6 e s v frequency C Useful relation 1 eq Photon energy Slide 5 Quick Example What is the frequency and energy of those Xrays in a beam Whose wavelength A 1 nm Answer 0 re v f q A Ehvl99eil Jl24KeV 3617 Hz 300000 THz Slide 6 Xray Spectra Electron volt 7 1 eV is the energy increase that an e ctron experiences When accelerated of l V le over a potential difference 7Or 1 eV qg AV l6el9 J Most of Xray applications in medical imaging are in range of 720 KeV lt E lt 150 KeV These energies correspond to frequencies in the range of 748el8 Hz lt v lt 36el9 HZ And wavelengths in the range of 7625e9 cm lt A lt 833e10 cm Un ltered Brems Characteristic 7 xrays Photon energy Filtering With high Z material absorbs low energy Xrays and shapes the output spectrum to reduce dose to patient 7 Characterrstrc Xrays Xray Lelectron knocked out In addrtron to brems X rays electron electron a tube usually generates characteristrc Xrays lament knocker Steps electron 7 Fast moving electron knocks off an inner orbital electron 7 Vacancy created is lled by a higher orbital electron 39 V I t f 7 The energy thus released is l mpac 9 highly quantized and shows up acceleranon as a sharp spectral peak VOItage Slide 8 Xray Interactions with Matter Primary effects at medical mm quotW energies 7 Photoelectric effect Energy of xray is absorbed by an orbital electron Electron is ejected from atom Electron from higher orbit replaces rele ejected electron asing photons 7 Scattering S T 7 p u q a Compton scatter 7 Some ofthe x ray energy transferred to an electron Xray travels on with an altered direction and less energy Coherent scatter 7 All of x gy interacts with the atom but radiated with arne energy in an arbitrary direction ener is later re Antiscatter grid 1 Slde9 Basic Xray Tube Design 1539 my me mm sums amen cowsc axs surth mums nvnrx Tuunsvsu mm ms msx mans wwnaw snmR Mmcmuis Rotating Anode Coolidge Type X7 Ray Tube Model RT 172 Circa 1936 Line Focus and Tungsten Target Gelterm Electric Company Slide 10 Mammogiaphy Tubes Slidell Schematic of 21 Diagnostic Xray System High Voltage Cable Transformer 250 V to 100 kV l Control U nit Main Supply 0 Cassette or digital detector E on o Slide 12 Scatter Reduction Scatter adds a low freq background across the image eld It lowers the contrast of structures of interest Antiscatter grids are used to remove Xrav Scatter scatter from the image 2 3 1 Z FD quot AntiScatter Grid Xray Efficiencies Xray generation is a fairly PVI inef 01ent process 7 Most ofthe electrical power ends up as heat at the anode e Ofthe generated xrays only 2 end up being absorbed at the detector Photon energy is related to acceleration volta e 10 KV will produce up to 10 KeV photons Total number of photons is related to cathode current 7 T ical currents are in few mA yp Absorbed in lm m 2 de M Dangers of X rays Xrays are ionizing radiation Will cause tissue damage l rad amount of Xray radiation that imparts 100 ergs of energy gram 0 tissue Frame of reference 7 Typical chest xray exposure 10 100 mmds 7 Background mdiation in upstate NY 200 7 300 mmds 7 Exposure of50 mds causes mdiation sickness 7 Dosages exceeding 200 mds can cause death 7 Geneml public should absorb less than 05 radsyear 7 Radiation Workers less than 01 radWeek and less than 5 mdsyear By June 1886 workers were cautioned ofthe dangers associated W Xrays MedicalXRays11 e Nuclear Medicine 4 Consumer Producls 3 7Terreslrlal 5 Cosmlc 8 Internal 11 X Ray Detectors MAMMOGRAPHY amp RADIOGRAPHY TODAY I x Ray FLUOROSCOPY TODAY Image Imensi er XRzy Electron ALL APPLICATIONS DIGITAL DETECTOR Detector XRay XRay W TV Pickup Tube ANALOG I ANALOG IMAGE DIGITAL IMAGE Slide 16 So how is an image actually formed Filte V Collimator Patient Antiscatter grid Scintillator A Detector Slide 17 Cranial XRay Slide 18 Film Cassette Scintillator XRAY SCINTI LLATOR FILM SCINTI LLATOR Enclosure Light tight Xray transparent FILM CASSETTES Slide 19 Digital Detectors Project started at GE Research in early 80 s as a spin off of LCD development Imager as well as specialized electronics developed Slide 20 Panel Structure DlFFUSE REFLECTOR I GUS SlLlCON ARRAY Glass Substrate Glass Suhglml Shde 21 Flat Panel Detector Pholodlode E Amorph on 5 Silicon rra Glass Substrate Xray Scintillator Csl Shale1 22 Amorphous Silicon Why amorphous silicon large seamless detectors can be made Largest detector for chest applications 50x50 cm substrate 41x41 cm active area manufacturing equipment driven by LCD industry Dif culties caused by amorphous silicon compared to singlecrystal silicon 100 to 1000 X increase in defects defects are longlived 100 sec mobility reduced by 1000 X affects switching speeds defects give rise to complicated lag and offset dark current behaviors Dif culties are overcome by calibration careful management of detector timing for all applications Slide 23 Pixel Layout 4 SCAN LINE LDATA LINE l39 FET ll PHOTODIODE VBIAS Slide 25 Measures of Image Quahty 0 Limiting Spatial Resolution L SR 7 The highest frequency that can be visualized 0 Modulation Transfer Function MTF Measures how the detector passes signal as a function of spatial frequency MT 1 Duarul LSR pa lal Frequency cyclesmm MTF Modulation at detector outEut Modulation at detector input Slide 26 How To Measure MTF IEdge Response Function Measurement method Test object placed on detector surface I tilted 153 degrees to detector axis m I measure MTF perpendicular to that axis Image of object made With standard spectrum and imaging geometry 7 7 Image linearized using inverse of characteristic ction Bar Phantom Edge spread function determined from data Within Difficult and subjective analysis RO Edge spread function numerically differentiated to obtain LSF MTF is calculated as modulus of Fourier transform of LSF MTF at frequency I obtained by averaging over f 001pitch Slide 27 Modulatlon Transfer Functlon LSR 7 Screenfrlm has LSR gtgt 20 lpmm corresponds to 25 mm pixel Dlglt l Imager Digital GE 100 mm pixel M Sources of MTF degradation T 7 Lateral spread of light in scintillator F limited by C51 needles increases With scintillator thickness 7 Lateral spread of secondary xrays not signi cant away from kedges of Cs and I FilmScreen 7 Sampling aperture ofpixel Spatial Frequency CyCleSmm sincfxasincfya If film s LSR is better than digital why do we see improved performance in digital Slide 28 Measures of Image QualityDQE Detective Quantum Efficiency DQE SNR2 at detector output SNR2 at deteCt0l OUtPUt DQE SNR2 at detector input Patient Dose SNR signaltonoise ratio 0 Measures transfer of both signal and noise Slide 29 The higher the DQE Digital Imager the higherthe SNR and the greater the probability of detection FilmScreen Spatial Frequency cyclesmm High DQE in lowtomid frequencies aids detection w 4 Slide 30 Object DQE and Noise Improved DQE 05 7 lt Better Image Same Dose Standard ll Same Image Half Dose gt SNR5 SNR 35 SNR 25 Slide 31 Scintillator Converts xrays to light We use evaporated CslTl as the scintillator lateral spread of light 560 nm 175 nm full width half maximum CslTl is one ofthe most efficient known scintillators evaporated on light sensor surfaceintimate contact with light sensors grows with needle structure needles act like light pipesprevent similarto CslNa scintillator used in image intensifiers emission spectrum is well matched to quantum efficiency of diodes Slide 32 NeedleStructured CsI Light is emitted isotropically Needle structure guides light like a fiberoptic t0 the photosensor I H Slide 33 Positioners 39Positioners determine how an x ray system looks Mammography Screening diagnostic and interventional procedures for breast imaging Only true Xray screening modality Image Review Shde 35 Sen ograp he 2000D Workstation XRay Mammography Until 2000 completely lm based Standard of care today in breast cancer detection All women over 40 should get a yearly mammogram done Sensitivity 65 to 70 Positive predictive value 10 7 20 Film Screen Mammography Reduces Breast Cancer Mortality by 2040 But There is Signi cant Room for Improvement lide 36 Digital Mammography 7 Introduced in 2000 Currently 4 digital mammography products on the market GE s product lOOum pixel pitch txSiCsl detector 7 Better small object visualization 7 20 Fewer Call Backs for Patients 7 Reduced Retakes 7 Dose Savings 7 Computer Aided Detection slide 37 LambertBeer Law dx 2 1ZIae l ldl Where 7 12 is the xray intensity at the Linear attenuation coemcrent measurement plane L E 7 In is the xray intensity at the 1 source plane 7 z is the distance between the source amp msmt planes 7 u is the attenuation coef cient Also known as Beer Beer39s law Lambert or Beer s Law 7 KHzI Intensrty can be given as a photon density photonm0 A I0 XD In general u is a function ofx this complicates matters Slide 38 Inhomogeneltle s amp LambertBeer Consider a multilayer object to be examined LambertBeer Law has to be applied to each segment With a constant p In the extreme With a continuously varying p the application of this law has to be as a line integral 12 1expyl 11 l l 1 Inexpyl 1 ILgt5Xph I Xray contrast De nition for present 7 Difference in the x ray IO 10 intensities emerging from adjacent regions Sometimes given as a 12 1 111 t age III Fum contrast 121711 X l 00 2 or as a normalized percentage I 7 contrast 2 1gtltlOO I 2 II Slide 40 Summary Quick historical review of Xrays was given Block diagrams key components de ned Brief discussion of xray scattering An Xray beam traversing through an object is attenuated by the exponential LambertBeer Law The product of the attenuation coefficient and the path length of the xray beam in such a target is critical in establishing detectability Next time XRay Metrics Slide 41 Homework Lecture 4 0 Slide 29 discusses the broad definition of Detective Quantum Efficiency Using Internet search engine give a functional algebraic expression of DQE 0 What is the relation of DQE to XRay exposure 0 What are common measurement methods for establishing the DQE Slide 42 Instructor Contact Information Badri Roysam Professor ofElectrical Computer amp Systems Engineering Of ce JEC 7010 Rensselaer Polytechnic Institute 11 0 83911 Street Troy New York 12180 Phone 518 2768067 Fax 518 27662612433 Email roysaercseggie u Website htt www iedur sab NethIeetjng ID for o lcampus students 1281136180 Secretary Betty Lawson JEC 7012 518 276 43525 lawsob Qiedu Rensselaer Slide 43 Instructor Contact Information Kai E Thomenius ChiefTechnologist Ultrasound amp Biomedical GE Global Research Imaging Technologies NiskayunaNew York 12309 Phone 518 3877233 Fax 518 3876170 Email thomeniu crdgecom thomenius ecse iedu Secretary TBD Rensselaer GE Global Research m e 44 BP4ED 4962ECSE 4962 InUoduc ontoSubsu aceSengng and Imaging Systems Lecture 20 Nuclear MedicinePET Kai Thomenius1 amp Badri Roysam2 1Chief Technologist Imaging Technologies General Electric Global Research Center 2Professor Rensselaer Polytechnic Institute GE Global Research Center for Subsurface imaging amp Sensing Recap Finished up on MRI scanners The way gradients generate the spatial info to form an image Frequency and phased encoding K space fMRI and magnetic resonance spectroscopy Associated clinical applications 0 Let s talk about Projectsll Nuclear MedicinePET For most part of this course our focus has been on imaging physical objects We have looked for features which interact with our probes Attenuation with Xeray Impedance mismatches in pulseeecho metho s Variations in proton density in MRI via magnetization vec ors Nuclear medicine amp PET are quite different As with fluorescence based methods they image concentrations of exogenous chemicals injected into the patient The observability of these chemicals is based on radioactivity Nuclear Medicine 0 Imaging is done by tracing the distribution of radiopharmaceuticals Within the body 0 Radionuclides or radioisotopes are atoms that undergo radioactive decay and emit radiation 0 In nuclear medicine we are interested in radionuclides that emit x rays or 39 gamma rays 0 A radiopharmaceutical is a radionuclide bound to a biological agent The Gamma Camera Basics of Gamma Camera Imaging mnphhnr39 l hrn rm um lamr mm H 1mm Hyaml I ulhmalm xi u Nuclear Imaging Instruments N uclear Medicine Imagers How does this work Radioisotopes are injected into the body A radioisotope can be 7 a pure element leg I431 which connects to Thyroid r a biological agent labeled with radioisotopes like MIBleTc99m All isotopes have a half life All isotopes are expelled from the body with an associated half life Nuclear Medicine provides physiological images ie the metabolic activity of the organs process the radiopharmaceutical and concentrate it in the target organs for imaging Normal v Bone Cancer Scan Detector or Scintillator 39 lNaIl Emits light H whenever hit by gamma A ill ray Amount of light is proportional to gamma 7 Electrical Signals energy level Photomultiplier Tubes read the light signals Collimator and translate them Into Radiation electrical signals from Patient Crosssection of an Anger Camera Shield Around Head Mounting Ring Collimator Core Sodium Iodide Crystal Photomultiplier Tubes weave Gamma Camera invented by Harold Anger in early 503 Nuclear Medicine Performance Metrics Typical performance Energy resolution 95 10 FWHM response Spatial resolution 32 38 mm Uniformity 2 4 CRYSTAL PLANE OF cc LLIMAYOR I 5 COLLIMATOR PLANE 0F CDLUMATGR POOL or RADIOACTMTY Resolution v Ef ciency Tradeoff smumpmm lexamm 7 Malalsmd Fanumu Emum Elmmn Inmpr Photomultiplier Tube PM Used to provide several orders of gain 103 Each is given a voltage higher Ihan me previous one e39 arrives with enough energy a ejec electrons Photomultiplier Tube PMT Advantages Disadvantages Microchannel Plate A 2d array of photo multiplier tubes Hv snmcamuumv We Wm Szwndsly ems Oulnul quotWW Emu channel evecllons Nuclear Medicine Images 0 Typicai image 7 64 by 64 pixeis 0 intensity gives counts per pixei 0 Pseudocoior often used 0 Nudear med imaging 7 MUGA muitkgoted acquisition 7 Whoie Body PECT Cardiac Study Cardiac Study 0 Evaluation of the coronary artery circulation Myocardial perfusion 0 3D Studies of the radionuclide activity SPECT Scanners 0 Single Photon Emission Computerized Tomography Store radionuclide emission data from multiple projections Projections taken every 3 or 6 degrees T Use CT type algorithms to determine the location and lt gt degree of accumulation of agent Clinical Applications In the 70 s amp 80 s SPECT was largely replaced by CT and MRI scans because they provided superior resolution Recently SPECT has returned to prominent use especially in diagnosing cardiac and neurological abnormalities While CT and MRI scans only provide images of static brain anatomy SPECT offers clues to brain function by tracing blood allocation Pros of SPECT Unlillte MRI and X ray ECT produces 3 D images that relate an organ s function This allows for better relay of extent of disease and reveals the course of the disease earlier Large amount of data on brain function already comes from SPECT scans Simple process with immediate results Much less expensive than MRI or PET 1000 per scan Covered by insurance when brain injury is present Cons ofSPECT 0 Unlike MRI and X ray there is an injection 0 Claustrophobia is a cause for concern 0 Quality of image can be lessened by patient movement PET Imaging 0What is PET Imaging A technique tracks biochemical and physiological processes in vivo 0 Uses tracer compounds labeled with positron emitting radionuclides 0 As such it is considered a form of functional imaging 0 Positron Emission Tomography oFunctional v Anatomical Imaging Choose X Ray Nuclear Medicine MRI CT US oHovv is PET different from Nuclear PET Positron Emission Tomography 0 Certain radionuclides emit positrons 0 When a positron meets an electron they annihilate eGCh Other v i Coincidence o This annihilation results in cm a generation of two gam ma rays 7 The gamma rays travel ih opposite directiohs 7 The ehergy of these gamma rays is 511 Kev 0 PET Imaging is based on detection of these gamma rays Radiaiion naiecioi PET Systems Event Detection 0 Several gammadetector rings surround the patient 0 When one of these detects a photon a detector opposite to it looks for a match 0 Time window for the search is few nanosecs o If such a coincidence is detected a line is drawn between the detectors 0 When done there will be areas of overlapping lines indicating regions 0 radioactivity LOR Line of Response Coincidence Events 0 Three Types True 0 The event we are after Scatter 0 At least one Com pton scatter event 0 Wrong LOR Random 0 Unlucky break How Does PET Compare With Other Imaging Modalities 0 PET provides images of molecular level physiology accumulation ofthe metabolically active radionuclide 0 Extends capabilities of other modalities Like CT it uses tomographic algorithms Like Nuclear images represent distributions of radiotracers 0 That39s where the similarity ends Here are three studies of a patient done with CT MRI and PET scanners CT Scan MRI Scan PET Scan Report Patient Deceased Report Normal Source Material httpwwwaehealthcarecomusenfun i mqnmedicinenmindex newhtml Siemens amp Philips web sites for nuclear medicine amp PET httpwwwcrumpuclaedusoftwarelppl pphomehtml httpthaverdartmouthedubpoaueENG G1671320Nuclear20Medicinepdf httpzootradioloqvwiscedubloclltbme 530lecturesL01lntroppt Summary 0 Introduction to Nuclear Medicine and PET imaging Additional examples of agents probes introduced to reveal subsurface phenomena Today s focus on radioactive labeling Review of instruments Relatively straightforward devices Signal to noise ratio challenges need to limit exposure 0 Powerful clinical tools 0 Much of today s research focused on PET and extensions of PET technology Homework Lecture 20 0 Using internet sources discuss the patient and clinician safety issues from the use of radioactive tracers in nuclear imaging Avalanche photodiodes lAPDsl are sometimes proposed as an alternative to photomultiplier tubes Discuss the pros and cons of these two types of devices Instructor Contact Information Professor of Electrical camputer amp sstenis Engineering 0 ice JEC 7u1u Rensseluer Puiytecnnic institute 11a 3 streettmv Nenvurk121su Phone 513i 27573057 Fax 513l2767626U2433 E H r g ec ergiedu website ntt www lEdLlm sub NetMeetirig ID for elfeumpus studentsl 1231135130 Secretary Lumine MicnueiidesJEc 7012 51m 27s 73525 micnuirgiedu Rem el 1 r Instructor Contoct Information Kai E Thomenius Chref Technotogrst Uttrosound amp Bromedrcot Of ce KWVCEOOA GE Gtobot Research Email nomenrucrd ge corn nomenrusecsergr edu Secretary Lororne Mrcnoehdes JEC 7012 518 276 78525 michot rgr edu GE Global Research m BMED 4962ECSE4962 Introduction to Subsurface Imaging Systems Welcome and Introduction Kai E Thomenius1 amp Badri Roysam2 1Chief Technologist Imaging Technologies General Electric Global Research Center ZProfessor Rensselaer Polytechnic Institute GE Global Research Center for Subsurface magmg amp Sensmg Basic Class Data BMED480001 ECSE480001 3 credits Instructors 7 Dr Kai Thomenius 7 Dr Badri Roysam Tuesdays amp Fridays 1230 150 Classroom JEC 4107 What this course is about 0 You can expect to learn the basics of imaging objects that are hidden under a surface in the context of realworld medical applications 0 Along the way we ll also learn about exciting applications requiring subsurface sensingimaging 7 Medical Imaging Systems 7 Bio imaging biology biotechnology Motivation for Medical Imaging Healthcare costs Enormous spend18T in us Admin casts 3005 shrinking population of doctors System cannot handle excess capacity Chronic care 70 aicost Huge markets m sswa imamangi Emader Pharmu Global Diagnosis Dmgnnshcs Healzhcnre lmuging Oppa nity Clinical ef cacy 50 of heart failure death l265kyr39l 323 usx healthcare spend on strokes I Adverse drug events 770K US deaths or injuries year 9 people 100 unintended infection Healthcare Trends Drive Imaging Growth mum Mindwiduuls as man Peopleavarlhe age was requlre a to Allmes as many Imuglng 3m mvestlgallons as those under me age a as mm 55 P am Aging Population 13 Globally Eurape 11 25M Japan 20 China 5 Japan us may Chum Cancer Heandlseese 1 killers ammmmme 7 same 13 ofull deaths chum We 7 1 Europe incidence en 7 Tun cause him US mm obesity an 12 Rlse 2quot 2quot 5 quot1M Mere pntlems WlUl Mnmpie Diseases Advent ofdlagnusticrdriven disease management Technology fusion shows anatomy molecular therapeutic efflomy From snape m Molecular Improved image quulnylorhetter dlngnoses Fusterexam nmes 5mm empeuem and nunemvaswe How can we look under surfaces Computer gt Image Decision Features Coordinates other Surface 1 Probes Detectors A Surface 2 Takes ideas from multiple disciplines working together What subsurface problems are we talking about Imaging 7 MRCTUltrasoundOptical Sensing 7 Decision making based on acquired data Characterization 7 Tissue elasticity Spatial registration 7 Procedure guidance eg biopsy mine location Examples of Imaging 3D4D Fetal imaging TSmaH Animai imaging Coronary Artery Visualizatio w MRl ulcerated plaque Examples of Sensing USMammngraphy Fusion Luggage Inspection Example of Positioning Re istration My Treatment Plan 3 Ultrasound Applicator Ablated Tissue M lmmy Example from Noninvasive Nonincisional Surgery Determination of Features of Target under Study cm mmmrm tummy x r mm Rummarm whmmy u mm z mums M1012 Tissue E ashc Properties Interaction between the probe and t t th nf ti Temhertz Sca amg arge conveys e 1 cum on Homeland Sammy Types of Probes For the purpose of this course the term Probes describes v u um mun Partially How Probes Interact with Media quot39 39 a quotI 39 such as a variation in propagation characteristics of the medium 5 2 s g E 9 3 n1 3 u 3 E 5 g 5 gt Propagationldiffraction D iffu39sion Scatte ri ng When can we see the subsurface object Image Decision Features t h Coordinates ot er Surface 1 Probes Detecm s v s Surface 2 Probe should reach the ob39ect Enough signal should reach the detector The detected signal should be affected by the object P Nr Example Example tissue but are attenuated by bones Medium Object w Absorptien I Scattering 7 1quot t Optical 3D Co foEal Microscope v Tumor Blmod Vessels Types of Probes in our Examples 92 39 Coronary Amery Visualization w MRI Ulcerated Plaque Cadaver Hean w CT Common Characteristic to Probes Wave Physics Subcellular Biology Ti 100nm10pm m Ultrasound Underwater Exploration Underground Diagnosis Radquot 71 10 Outline of Course Topics THE BIG PICTURE What is subsurface sensing amp imaging Why a course on this topic EXAMPLE THROUGH TRANSMISSION SENSING XRay Imaging Computer Tomography COMMON FUNDAMENTALS propagation of waves interaction of waves with targets of interest PULSE ECHO METHODS Examples OPTICAL IMAGING MRI A different sensing modality from the others Basics of MRI MOLECULAR IMAGING What is it PET amp Radionuclide Imaging IMAGE PROCESSING What We Expect From Students Prere quisites 7 ECSE2100 Fields amp Waves 1 7 ECSE2410 Signals and Systems Willingness to learn some Biology Medical Terms Physics Math Computing basics of programming Prior exposure to computers Internet and Course content reasonably exible 7 Expectations assignments etc proportional to prior background 7 Choice of topics and coverage to meet student interests Grading Assignments 7 30 7 To be handed in within one week 3 Quizzes 7 30 Project 7 40 teams up to 2 ok 7 lndepth review of a specific course topic eg processing of raw data sets 7 Topic subject to instructor approval All students should show independent homework and projects Generous reward for effort evidence should be shown Study Materials Lecture notes Other class handouts Selected internet sites MATLAB We have campus licenses at Rensselaer UPRM NEU amp BU Summary Subsurface imaging 7 Course Focus 7 Many applications to Biology Medicine 7 Probes media and probemedia interactions Methods by which we examine contenm of hidden volumes 7 Wave theory the common thread Common to electromagnetic and acoustic probing Next Class 7 Overview Projection Imaging CenSSIS Major Research Center on Campus lVLission 7 Create a uni ed engineering discipline for sensing and imaging objecm that are hidden under surfaces 39 quotszrsz Pmblems Similar Solutions Bene ts 7 Learn from similarities It s all the same physics 7 Learn from differences Why is an Xray beam different from ultrasound WWWecseIpieducenssis 7 Learn from experience With other probes GE Global Research Niskayuna 7 Advanced Research for all of GE 7 Founded in 1900 in a barn in Schenectady by Steinmetz Imaging Products amp Services 7 GE Medical Systems CT Functional Imaging MRI Ultrasound XRay Support PACS Workstations GE URLs 7 wwwgecom 7 wwwgehealthcarecom Instructor Contact Information Badri Roysam Professor of Electrical Computer amp Systems Engineering Of ce JEC 7010 Rensselaer Polytechnic Institute 110 8 11 Street Troy New York 12180 Phone 518 2768067 Fax 518 27662612433 Email roysamecse giedu Website htt WWW 39edu7ro sab NethIeeting ID for of campus students 1281136180 Secretary Laraine Michaelides JEC 7012 518 276 78525 rm Bensselaer n mquot quotu mun Instructor Contact Information Kai E Thomenius ChiefTeclmologist Ultrasound amp Biomedical Of ce KWC300A GE Global Research Imaging Technologies Niskayuna New York 12309 Phone 518 3877233 Fax 518 3876170 Email Lhomeniu crdgecom Lhomenius ecse giedu Secretary Laraine Michaelides JEC 7012 518 276 78525 W GE Global Research BMED4962ECSE4962 Introduction to Subsurface Imaging Systems Lecture 6 CT Scanning II Kai E Thomenius1 amp Badri Roysam2 1Chief Technologist Imaging Technologies General Electric Global Research Center 2Professor Rensselaer Polytechnic Institute GE Glubal Research Center for Subsurface Imaging amp Sensing Outline of Course Topics 0 THE BIG PICTURE o PULSE ECHO METHODS What is subsurface sensing amp Exummes imaging MRI Why a course on this topic EXAMPLE THROUGH tAhiilif5Psensmg modality from TRANSMISSION SENSING XRuy Imaging BaSIcs ofMRI Computerromogmphy 0 MOLECULAR IMAGING 0 INTRO INTO OPTICALIMAGING Whatis It AND SENSING PET amp Radionuclide Imaging 0 COMMON FUNDAMENTALS o IMAGE PROCESSING amp CAD propagation of waves interaction ofwaves with targets ofinterest Summary 0 Introduction to CT scanners o Emphasis on reconstruction algorithms Basic problem of reconstruction from projections Simple reconstruction using linear equations Fourier Slice Theorem Backprojection amp filtered backprojection Quick X Ray amp CT Review Measuring line integrals of attenuation coefficients u A av beam unveiling along Line J intensities I I 1 ex 7 39 WW The photon nuensmes 1 are prepmcessed io pmjemon data p e my hie I 71414 P IO Thus we lime to solve a set of integral equations for the summon cne icimts u quotY dutcmi Source 39 quot4 quot h 121203 Lauritschndf Quick X Roy amp CT Review Properties of the attenuation coef cient p The object patient is described by the spatial distribution on the attenuation coef cient M The attenuation coef cient i477 uF p ZzE is a function of de 39tv p mi material spect c Z atonnc number E energy of xray beam Usually the dependence of the beam energy E is only roughly considered by a beam hardening correction of the intensities L and E is set to an effective energy E A Em Source 39 quot quot 39 121203 Lauritschpdf Sinogrom Definition Sinogram 9 The representation of the projection data ppt 9 Radon transfoun 1R 1 p in a pBdiagrain is Called sinograin A point in spatial domain appears as a sinusoidal curve in the sinogram Hounsfield Units CT numbers of tissue in Houns eld units HU 3mm so Spleen mam s 40 Bane x nlel my ad rm am I mm m I Aduu39ll ma Liver Jan Jim CT number xii1W 1 000 MM Getting into CT Details Relation of Radon Transform Fourier Slice Theorem Filtered Backprojection All of these deal with making an image of ylxy from projections given by Pt6 where P N CD l xay Y is the projection of of Way along L is the distance ofL from the origin is the angle ofL Wit to the yaxis Radon Transform o In 1917 Johann Radon published an exact solution to this problem given below 0 None of the CT pioneers leg Hounsfield Cormack were aware of Radon s work nor is the direct expression being used in any commercial scanners o This expression includes a derivative of Plt given noisy x ray data this would make reconstructions challenging o Other problems include the discontinuity at t t this actually goes away if you do the Stintegration first 0 I m not sure which expression Matlab implements in their Radon inverse transform but it works amp is great for homework Mm 12Jid6jdt 1 5Pt39 27 t t at39 where t x cos y sin9 Fourier amp Filtered Backprojection 0 The most com mon reconstruction algorithm is the filtered backprojection method 0 Overview FBP Reconstruction steps Collectone projection immediately start processing the data Take its 1D Fourier Transform Multiply by filter response ramp filter or one of the variants Inverse 1D Take inverse 1D Fourier Transform Fourier Backproject across image space transform Repeat for all projection angles 1D Fourier transform collect Pro le in projection repeat for all projections Multiply by lter backproject across image space Excellent reference Kak amp Slaney Principles of Computerized Tomographic Imaging IEEE Press Ch 3 Motlob amp CT Reconstruction Matlab offers a forward radon and inverse iradon mefiles in our CT context R radonltheta oi contains an intensity image oTheta is a vector of projection angles 0R will give the resulting projections when plotted as a 2D image it will be a sinogram 7 iradonRtheta 0R is the projection data this is the point where we us a ly s art oTheta is the same vector as above Motlob amp CT Reconstruction o in addition to the radon m les Matlab glives you a SheppLogan Phantom m e 7 F39 phantummizt 7 lrnShEIMF39 0 With those two statements we get this gray scale SheppLogan image 0 Vge can now take the radon transform of t IS39 7 theta u 179 pruiectiun angles 7 R Xp redunm theta We can now display the resulting sinogra 7 figure imagesc hetaS XpRS culurrnamhut culur ar 7 xlabel theta ylabel mums Let s check how much worse things get with going back with the iradon mfile 7 P2 iradun 2 Fourier amp Filtered Bockprojection Lost time we discussed the Fourier Slice Theorem 1D FT of each projection gives one rodiol line ofthe finol imoges 2D FT oThe relotion to FBP con be m y pm Fm understood from t I Edch projection is nearly independent of each other 0 Can be more easily visualized in the frequency space 0 Only common point is at DC Thus we con odd their contributions to each other u freannzy damnin Relationship between spatial Fourier and Radon domain 2D Fourier transform 2D Radon transform radial 1D Fourier transform Fourier amp Filtered Bockprojection Before we can add the individual contributions ofthe Fourier transform we have to account forthe nonruniforrn sampling The ideal filterforthis is the pie ped vve ges shown in Figure lal Unfiltered processing would simply use response bl To emulate response al people have used the filter response c if we assume there are K projections over 180 degrees then the width of the wedge will i Wm be 27mm This is a The ramp lter shown in C l is 39 V 7 r r 77 x 7 intended to model this response S 7 K r J V Fourier amp Filtered Bockprojection The final reconstruction is the addition of the 2D inverse FT of each vveighted projection This step is called a backprojection since in effect it is u a 7 a smearing of each filtered quotquot projection overthe image plane There are several advantages of u w this method overthe Radon orZD D FFT methods 7 The process can be started D 0 immediately after the first projection data is in Any required interpolation can be done as an of the backprojection process Fou rier amp Filtered Bockproiectlon Finally the filtering process 7 7 can be adjusted to bring 1 Fw Klwl g about any modifications in quot39 the data The attached figure shows several such filters that have been used for this purpose Two examples are shown One can reduce higher frequency components by introducing a lower cutoff see Hann or Parzen filters Frequency sznal a Lak shewLogan Hamming 9 7 g i if i g xkx m n Frequency Freqst J 1 Getting Perfect CT Data 0 Most of the code in today s scanners does not deal with reconstruction rather data optimization 0 There is a real need to improve the quality of the raw data Corrections for detector sensitivity variation 0 Remember 9 ux lnlIDI Our measurements are critically dependent on uniform responses from all the detectors Correction for beam propagation effects and detector failings As it turns out doubling the path length seldom doubles the attenuation things are actually worse Beam hardening lower frequency xrays preferentially attenuated causing the remaining beam to have higher average ene g Scattered radiation Detector nonlinearities each detector has its own nonlinearities Some Web Resource httpwwwonidorsted uforidondpreprintsfb M motlob code for filtered bockprojection Designer Shepp Logon phontom Filter design possibilities Modified code available from httpwwwecserpiedu censsisSSICourse Some Web Resources oft esft itdtudllt ubkursus3l655 deLOS severol demos involving projections reconstructions etc ct imaging in general and projections in particular Very nice demos for l Resel calm Some Web Resources 0 ttpWwwctsimorg open source CT simulator 0 very nice tool to play games with various reconstruction parameters Has to be downloaded to run on your PC Completely menu driven app CT Scanner Evolution 0 First Generation 1970 Parallel beam design Onetwo detectors Translationrotation 0 2nd Generation 1972 Small fan beam Translationrotation Larger no of detectors 0 3rd Generation 1976 Multiple detectors Large fan beam 0 4th Generation 1978 Detector ring Source rotation Large fan beam Fan Beam Data acquisition in fan beam geometry Commercial scanner of the third generation acquire data in 391 fan beam geometry Reconstruction methods for fan beam geometry Rebiiniing r Resanipling offan beam to parallel beam requires additional interpolations 7 Application of reconstruction for parallel beam requires waiting time for acquisition ofinany projections Refounulation of the inverse Radon transform by a coordinate transfonnation from parallel to fan beam geometry Direct inversion CT Scanner Evolution Multislice scanners HelicallGElorspiral i Siemens scanners r Simultaneous Source rotation Toble Translation Data Acquisition Electron Beam Tomography Singlerslice Cr Quadrslice CT One my lube and m vuw ui detectors pmvldE 1 charmsu spaiiai data autism detectors in a single YDWaYE One way lube and multiple YDWS ui detectors pmvldE A channels uispaiiai data Manylhuusands ui detectors lquot a 20 array IVIulti slice CT Going beyond cone mum beam Area of keenest competition today Manufacturers are adding slices today 1 patient coordinate svslem 3 a we are at 64 anxious Next year who movement kn ows Multi slice CT Dates 1970 1978 m neeme beam r vzpaman an beam r5 smgle detector 199 future 0 mumfan beam wne beam I area ueieanr 2003 03 05 odf New Challenges vv Helical CT whiqu a In conventional CT quot 39 l quot 360 degrees of data are acquired for one image To get another image the gantry is moved to next location Helical CT covers a non planar geometry Patient Table moves anywhere from 1 10 mmsec Mostalgorithms are the same as with conventional An added interpolation step z interpolation is required Homework 6o 0 You will be supplied with o sinogrom of on unknown imoge lunknownhwk6dl Using Motlob s irodon function find out what the image is using theto 0179 Form sinogroms of the image with the radon function and varying numbers of projections ie different lengths of theto What happens to image quality with shorter vectors At what point would you soy the image quality is acceptable Homework 6b extra credit 0 Using the Forinodo filtered bockprojection code change filter parameters for 0 lower bondposs Demonstrote loss of spatial resolution Summary Several details involving CT scanner operation were reviewed Distinctions among the major methods for image reconstruction In particular the role of the convolution filter in FBP was considered CT related resources on the web were identified The various generations of CT scanners were defined MRIUS Guys View of CT Instructor Contact Information Badri Roysam Professor of EIectricaI Computer amp Systems Engineering Office JEC 7010 RensseIaer Poiytecnnic Institute 110 8 h Street Troy New York 12180 Phone 518 27678067 Fax 518 27662612433 Email roysam Qecsergiedu Website nttg www rgi eduNroysab NetMeeting ID for offcampus students 128 113 61 80 Secretary Loraine Michaeiides JEC 7012 518 276 78525 micnairgi edu Instructor Contact Information Kai E Thomenius Chief Tecnnoiogist UItrasound amp Biomedicai Office KWVCEOOA GE Giobai Research Imaging Tecnnoiogies Niskayuna New Vork 12309 Phone 51813877233 Fax 51813876170 Email tnorneniucrd ge corn tnorneniusecseirgi edu Secretary Loraine Michaeiides JEC 7012 518 276 78525 micnairgi edu Rensselaer GE Global Research m i BMED4962ECSE4962 Introduction to Subsurface Sensing and Imaging Systems Lecture 24 Multispectral Imaging Kai Thomenius1 amp Badri Roysam2 1Chief Technologist Imaging Technologies General Electric Global Research Center 2Professor Rensselaer Polytechnic Institute GE Global Research Center for Subsurface lmagrng amp Sensrng Recap Molecular Imaging image the spatial distribution of one or more specifically chosen substances The chosen substances are known to be informative about biological processes of interest Eg apoptosis angiogenesis metabolic activity Lots of molecules of interest Impractical to build a custom imaging system for each substance of interest Contrast agents a practical strategy Attach specifically to molecules of interest Capable of revealing themselves using existing imaging modalities Increasingly available for most imaging modalities Extrinsic amp intrinsic contrast agents Multiplexing Use of multiple contrast agents at once Capable of providing molecular associations Recap Molecules amp Spectra Abiornllun W Fluoreicerwe Excitation E mission Relallv Internally Jan mu m we 1m Wavelength nm Absorption amp Emission Spectra a The emission spectrum is a more iiS ii n More on spectra fingerprints today 7 Fluorescence Phosphorescence amp Bioluminescence phenomena 7 Immunorluorescence The use of lluorescently coniugated antibodies to tag speci ic proteins of interest 7 Multi lexing simultaneous use of multiple tags to image multiple molecules preserVing relative Motivation for Spectral Imaging Multiplexing mage more than one molecule of interest at once Spectral discrimination Reso Ive two or more molecules based on the overall shapes of their spectra rather than 1 2 readings A way to pick up more subtle differences between molecules Approaches to Building Spectral Imaging Systems Detectors MOSt Spem c Narrowband Wideband arrowband Excitation eg laser A A Wideband eg White light IA quot Spectral channels LeaSt SpeCi c Basic fact To achieve more molecular specificity we must use instrumentation with greater spectral selectivityresolution 11 Multi and Hyperspectral Imaging Multispectral imaging Record intensities at more than one wavelength at each pixel 2 50 channelsbands Gives us an array of intensities at each pixel instead ofjust one value e can be clever in selecting the bands and their widths Hyperspectral imaging Record an entire spectrum at each pixel Could be emission or absorption spectrum Dense amp nonspecific sampling across the spe rum Usually uniform sampling of spectrum 50 or more channelsbands An array of wavelen th indexed intensity values instead OfJUS one value at each 39er spectral finger print llllllllllllllllll Multispectral Example Blood Oxygenation E I 10 E At 570nm the absorption of i 12 39 quot3 HbO2 and Hb are the same a 8 use this band as a g 0 5 reference g At 600nm the absorption of 395 HbO2 and Hb differ N4 5 5amp0 530 650 6amp0 times use this band as a Wavelength nm test Absorption spectra of HbO2 I and Hb O Hb Hemoglobin Dr J Beach Application to Functional Retinal Imaging v E n L mm 5 mm N A E HINDUS AMEKA xrr FUN L Optical train of the image splitter The Optical Density Ratio Trace vessels Identify interior and exterior regions Calculate the ratio of minimum reflectance on the interior and average in the exterior neighborhood ll Calculate the ratio ofthe 35 I a OD s at 600nm and I 570nm Ostu 1m 10g WEE This is negatively correlated Izmreg with blood oxygenation ODR ODsooabxm W OD 570 chm Retinal Example contd R99quot air breathing Pure 12 bireathingV pquot r M V Dl A 3219 ooh OD ratio is negatively correlated with blood oxygenation Why Build Hyperspectral Systems Multispectral systems are optimal for a given application with known spectra Such as the retinal example we saw Hyperspectral systems offer exibility across applications They may not be optimal for a specific application Often useful as a prototyping tool for building multispectral systems Prisms Filter Wheels amp Diffraction Gratings G Narrows dawn the spectral band Array of detectors amp Splits a beam of light bywavelengths Pushbroom Scanners One rovof39 of image region y l PrismIGrating Linear Array Pushbroom httpIwwweomonlinecom Wideband Excitation Example Sunlight The hyperspectral image Is also referred to as a 1 cube39 Each pixel stores an entire spectrum Re ectance w Wavelength Micmu pm Courtesy Prof Miguel Velez UPRM The Zeiss META Microscope The fluorescence light after the pinhole is passed through a grating to an array of 32 detectors PMT s Produces a lambda stackquot Ix y z 1 A 32point emission spectrum at each pixel 5Label Confocal Imaging Example s Microglia Iba1 Excitation spectra i Emission spectra LNEK 5Label Confocal Imaging Example Microglia Iba1 Emission spectra Dealing with Overlapping Spectra mmmv eference Spectra Rimmed 500 le 520 5m an 551 sec 570 55a mum wavelength mu Ron ROI Nucleus histone GFP Fusion Actin laments uorescein conjugated phalloidin The peaks are separated by only 7nm Dealing with Overlapping Spectra Reference spectra R11amp R2 1 Spectrum at any pixel is a linear mixture of the reference spectra Sxy 21 141065 ZgtltR1l 142 x y Z X R2 1 Unmixing Resultquot A and A quotAbundancesquot UnmiXing Problem Compute A1 and A 2 at each pixel subject to constraint A1 A 2 1 SimpleMinded Unmixing Algorithm Define a goodness of fit measure Ex y z Z Sx y z 1 A1x y 2R11 A2x y 2R212 A At each pixel compute E for all possible values of abundances A1xyz 01 in steps of A2xy21 A1xyz Choose the value that minmizes E Dealing with more than 2 abundances A unit simplex a set of numbers that are each in the range 01 and which add up to 1 N 241 431 There are only N1 independent variables Unit Simplex Example N 3 Let u1 u2 be two real numbers 1 1 02 1e 1 1e 2 0391 A1201 A2 l a39lgtltcr2 A3 1 0391gtlt1 0392 Whatis A1A2 A3 Tree Diagram Summary Spectrally resolved imaging Essential for multiplexing Valuable for resolving molecules Multi and hyperspectral imaging Specialty vs general purpose systems Spectra unmixing algorithms Homework Lecture 24 Download the le 5Iabel spectra 39om the course website These are the 39 39 i MET 39 the 5Iabel cl ass metcymm 1 Usmg MATLAB or Exee1 create a synthelm m1xtme loe GFAP and CyQuanlspecIraw1lh abundance vanes 01 20 and 30 yespeetwew 3Repea12 by vanmg the mmng 731105177 5 moremenls 5 10 15 etc Instructor Contact Information Badn Roysam Professor of E1ectncaL Computer amp Systems Engmeenng Of ce JEC 7010 Rensse1aer Pomechmc mettute 1118m street Troy New York12180 Phone 518 27678067 Fax 518 276762612433 Em Website mtgMW 1edur0 Gab NetMeeting ID for offcampus students 128113 61 80 Secretary Larame M1chaehde5m1cham1edu 518r27678525 Rensselater Instructor Contact Information Kai E Thomenius Chief Technologist Ultrasound amp Biomedical Of ce KVV C300A GE Global Research Fax 518 3876170 Email thomeniucrdgecom thomenius ecser iedu Secretary Laraine Michaelides michamiedu 5182768525 GE Global Research m Narrowband Excitation Example lasers 350 xyz Finegrained in xyz Coarsegrained in A V DAPI DNA Stain LeX t GFAP Tourquoise tubes and Lewis X blobs Laminin stains vesse s bulbs BMED4962ECSE4962 Introduction to Subsurface Imaging Systems Lecture 5 Xray Imaging cont Kai E Thomenius1 amp Badri Roysam2 1Chief Technologist Imaging Technologies General Electric Global Research Center 2Professor Rensselaer Polytechnic Institute GE Glubal Research Center for smsmace imaging amp Sensing Review of Last Lecture Quick historical review of X rays was given Block diagrams key components defined Brief discussion of x ray scattering An X ray beam traversing through an object is attenuated by the exponential Lambert Beer Law The product of the attenuation coefficient and the path length of the x ray beam in such a target is critical in establishing detectability Today Digital Detectors X ray Metrics Outline of Course Topics THE BIG PICTURE PULSE ECHO METHODS What is subsurface imaging Examples Why a course on this topic MRI EXAMPLE Projection Imaging A different sensing modality xRaymaging from the others Computer Tomography Basics of MRI 39 COMMON FUNDAMENTALS 39 MOLECULAR IMAGING Propagation ofwaves What iS it Interaction of waves with PET amp Radionucnde Imaging Iargeis orimereSI 39 IMAGE PROCESSING amp CAD COM 39 PACS Human Size Ef ciency Digitlzauun Data delivery Radiologist rzlnls Resuiullnn Prepmcessing Datadisplay Physlcian m slnra e Experience Con guration Wurkflow Condltlnn kVp mAs Rest Tube ltration Exam type Scanergrid Postprocesslng Da DOE Coiiimation ESE dose wwwum eggmemagsmsiggzmsysmzmumEa Digital Detector Front End V t H 39iamp De Digital Pixel Matrix Transmitted xrays through patient Dig a process 1 A Analog to Digital lel39SlOl39l Charge Xray converter collection xrays gt electrons device wwwum aglmeetmgslamasZgdm65959783142414 Ea Detector Detoils TFT Csl phosphor swam x hwsphur Csll dizcent late line Charge thudinde Storage capsular Lil39 lts to electroni lo light to wwwum agymetmgslamasZydm 597831427414 Ea Selenium based Detector D t cir e or aSe TFT array Incidem Hays Tn marina in Staten charge rays to electrons to electronic signal Performance Metrics Sig no l to noise Rotio SN R SNR determines tne detectabiiity of an obiect Signai derived from gtlt7ray quanta Noise comes from a variety of sources 7 X7 a quantum statistics Poisson distri ution Signai processing steps criticaito image quaiity 7 Correction for detector variabiiity efects 7 Postprocessfiitering Quantum noise 0 For a digitai gtlt7ray detector my aquot system Witn square pixeis i iftne average number of gtlt7rays C a J I a recorded in each pixei is N a a t Pholunsnisuibuinn in v Handnm Pinen 7 then the noise per pixei Wiii be My o Statisticaidistribution associated Witn gtlt7rays is tne Poisson distribution reiation faiis out I I Skamm De 7 The above 9 directiy from this fact smart DevFaTon 0 pm um as name 1000 Photon Avmgu Par mu m n a Poisson Distribution Poisson Distribution is a probability distribution given by fk 1 W k ifthe expected no of occurrences in indi Li a space is 7L then the probabiiity quot that there are exactiy ilt occurrences n r is given by fikJJ in Wt s in p Signaltonoise ratio The signaietoenoise ratio SNRi is given by N N SNR a39 m 7 N When the number of xerays N is increased the radiation dose aiso increases To dou bie the SNR the dose to the patient needs to be increased by a fact Contrastetoenoise ratio CNRi for any two intensities izi at a detector is give iland IIrIzLIIrIzI CN39Rr r a39 Here N is the nominal value of photons reaching the detector SNR and CNR Background 4203 i33 Object 4118 i33 SNR1Z74 33 26 CNR 4203 4118 33 Visual Detection of Object SNR CNR is x ray quanta dependent Detection is determined by CNR and object size C D phantom gt quot k SNR gtlt d x C 239 u I C contrast d diameter k 3 to 5 for detection Contrast Other Measures of Image Quality Limiting Spatial Resolution lLSR The highest frequency that can be visualized Modulation Transfer Function lMTF Measures how the detector passes signal as a function of spatial frequency MTF 10 x LSR Spatial Frequency cyclesmm Modulation at detector output MTF Modulation at detector input MTF1 Source 2036425139infodownloadetclbreastxl939304ppt MTFO5 ll 0 Modulation Transfer Function MTFlI Spatial resolution An imaging system s ability to render the contrast of an object as a function of object detail mumnm Illlllllllmmmm Modulation Transfer Function OLSR 7 Screeniilm has LSR gtgt 20 lprnrn corresponds to 25pm pixel Digital lGE 100 um pixel oSources of MTF degradation 7 Lateral spread of light in scintillator limited by Csl needles lmscreen increaseswitn scintillatortnickness 7 Lateral spread of secondary gtlt7rays not significant away from k7edges of C5 and 7 Sampling aperture of pixel Spatial Frequency cyclesmm sincltgtltalsincltyal Iffilm s LSR is better than digital why do we see improved performance in digital MTF For Direct Indirect and Screen Film Streen7film Min R2000 Kodak Spaiial Resolution IDmm Measures of Image Quality DOE Detective Quantum Efficiency DQE SNR2 at detector output SNR2 at detedor OUtPUt DQE SN R2 at detector input Patient Dose 0 SNR gives the transfer function of both signal and noise The higherthe DQE the higherthe SNR and the greater the probability of detection FilmScreen Spatial Frequency cyclesmm High DQE in low to mid frequencies aids detection DQE Definition 52 MTF3f NPS f X X C DQEiX Where f is the spatial frequency lpmm X is the exposure mR and S Median Signal Level cts ie amplitude ofinformation MTF Modulation Transfer Function NPS Noise Power Spectrum ctsAZ mmAZ C Incident Xray Fluence Xrays mmAZ mR DQE describes the measured SNR in relation to an ideal detector SNR2 is deduced from the ratio ofMTFAZ signalAZ to the NPS noiseAZ Detective Quantum Efficiency Radiography 05 10 15 20 Spatial Frequency cyclesmm Calibration of Digital Detector Dark Image I Offset Diode leakage FET charge retention Electronic noise Calibration of Digital Detector o Offset Corrected Dark Image 0 Electronic Noise Calibration of Digital Detector Offset Calibrated oAmplifier gain variation Pixel to pixel gain variation Calibration of Digital Detector Offset and Gain calibrated Flood exposed image Poisson statistical x ray noise Electronic noise Apply Corrections Low dose before and After Offset Correction 0 High dose Apply ZCQWeCtiH iS Advanced Applications 0 Tomosynthesis 3D X ray 3D Breast Imaging Tomosynthesis 3D imaging addresses the major problem with mammography today superimposed tissue 3D imaging may enable compression reduction 7 Tissue immobilization vs compression 7 Compliance with screening protocols Single tomo exam in MLO position may replace conventional mammography potentially enabling dose reduction Tomosynthesis Concept Prototype System Porometers Prototype based on GEMS Senogrophe DMR Revolution flot ponel detector motorized tube motion ossembly 11 projections over 25 degrees 0 75 sec potient exom time Totol dose 15x 0 single mommogrophic view O75gtlt 0 standard mommogrophic screening exam 0 100 micron pixels 0 1 mm l3dl slice seporotion Tomosynthesis Gaal Limited 3D reconstruction to remove overlyingunderlying structure All image planes visualized using a single acquisition Acguisitian 0 Vertical tube motion Total tube angle 5 15 Number of Projected Images 15 25 Exam length 5 10 sec lsingle breathholdl Slice thickness 1 cm Enabled by GE RevolutionTM detector Small Changes to Rad System allows for 3D Imaging Image Reconstruction in Tomo Data incompleteness From a CT perspective data is very sparse Limited angular range z resolution Insufficient angular sampling streaks Truncated projections inconsistency Reconstruction Concept Shift and Add Vertical slice Reconstruction tnrougn object ofsingle plane Reconstruction of vertical slice through object AIAA Projections at different angles Add 41 J u IA I 39 Shift I Artifacts quot quot r Tr asN39 r39 I of projections Contrast I quotblurringquot of artifacts depends on N projection anglesl tube trajectory etc An Example Stondord 2D x ray images courtesy of Dr Dan Kupans MGH Tomosynthesis Missed Cancer Standard Mammogram Tome Slice MLO Spiculated Lesion Tomosynthesis Images courtesy of Dr Dan Kopans MGH An Example 3D Tomosynthesis Images courtesy of Dr Dan Kopans MGH Rod Tomo Example Low Dose 3D Imaging Receiver Operating Characteristics Receiver Operating Characteristic ROC curves Most basic task of the diagnostician is to separate abnormal subjects from normal subjects In many cases there is significant overlap in terms of the appearance of the image Some abnormal patients have normal looking films Some normal patients have abnormal looking films ROC curves are a tool for assessing the performance of a hypothesis testing algorithms 2 x 2 Decision Matrix Actually Abnormal Actually Normal Diagnosed as True Positive False Positive Abnormal lTPl lFPl Diagnosed as False True Negative Normal Negative lFNl lTNl ROC curves cont 0 Fora single threshold value and the population being studied a single value for TP TN FP and FN can be computed o The sum TP TN FP FN will be equal to the total number of normals and abnormals in the study population 0 True diagnosis must be determined independently based on biopsy confirmation long term patient follow up etc ROC Curve decision ihreshuid true positive fraction of patients o 2 ti A decision criterion false positive fraction 3 Summary 0 Design ofdigital x ray detectors was described 0 Performance metrics MTF DQE for x ray performance were given Justification for digital detectors was based on these 0 Tomosynthesis concept introduced 0 Brief review of ROC methods for hypothesis testing was given 0 Next time Introduction to CT Scanners Homework Using web resources or sources given below describe the key steps of the direct conversion process with amorphous Selenium How are x rays converted to electrons What is the relative performance MTF or DOE with respect to the CsI Photodiode approach Which would you buy and why Tl h vwwv dnndirk in ca 39 39 ballerina of Direct Digital Rodiologypdf r Hoheisel et ol quotModulation transfer function of o seleniumebosed digital mammography system IEEE Proc Nuclear Science Symposium 2004 358973593 Instructor Contact Information Eudri Roysum Professoer Electrical Computen sisstems Engineering Office JEC 7010 Rensselaer Polytechnic institute 110 8 Street Troy New Vork 12180 Phone 51827678067 Fux 518276762612433 Website httg www mi edulroysab NetMeeting ID for D rcumpus students 128 113 61 80 Secretary Betty Lawson JEC 7012 518 276 78525 lawsob ml edu Instructor Contact Information Kai EThomenius Chwef Techno ogwst Uhrosound amp Bromedwco Of ce KWiCZOOA GE G obo Research mogwng Techno ogwes Niskoyuno New York12309 Phone 518 38777233 Fax 518 38776170 Email thomemu crd ge com thomemus ecser edu Secretary TBD Rens elaer GE Global Research BM ED4962ECSE4962 Introduction to Subsurface Imaging Systems Lecture 10 Propagation of Waves III amp Ultrasound Kai E Thomenius1 amp Badri Roysam2 1Chief Technologist Imaging Technologies General Electric Global Research Center 2Professor Rensselaer Polytechnic Institute GE Glubal Research Center for Subsurface Imaging amp Sensing Outline of Course Topics PULSE ECHO METHODS Examples THE BIG PICTURE MRI What IS EUbSUI face senSing 8 A different sensing modality Imaglng from the others Why a course on this topic Basics ofMRl EXAMPLES THROUGH MOLECULAR IMAGING TRANSMISSION SENSING Whams it X39Ray Imaging PET amp Radionuclide Imaging ComputerTomography IMAGE PROCESSING amp Intro into Optical Imaging CAD 0 COMMON FUNDAMENTALS interaction of waves with targets of interest Recap o Propagation Models Fresnel or near field Fraunhofer or farfield Near to far field transition distance 0 Starting w Rayleigh Sommerfeld we derived a simplified CW far field expression for the probe field Fourier transform of the aperture 0 O Neil s expression fora circular aperture 0 Today We have some concepts on wave propagation We will talk a bit about two key wave related concepts Let us take some of that info and design a scanner Ughtpropaga onPar des orVVaves Isaac Newton A particle or corpuscle guy Christiaan Huygens A wave guy introduced the Huygens Principle 0 Thomas Young 1801 Double slit experiment validated the wave concept 0 Actually both concepts work and are useful in understanding light propagation Huygens principle Huygen39s principle offers an explanation for wh and h waves bend or dlffract when passing an obstruction every point on a wave front acts as a source of tiny spherical wavelets thattravel forward with the same speed as the wave the wave front at a later time is then the linear superposition of allthewavelets Plane wave iron at sound ZDDBnateschzs 28 ah SlEal DDtlES aat I A Young 5 double slit experiment V v 0 ln 1801 Thomas Voung performed an experiment that irrefutany demonstrated the wave nature 0 li t 7 before thisthere had been a lat of debate between the particle camp Newton and the ave camp Huygens o Monochromatic single frequency light is rst shone through a single slit 7 this makes the lightthat passesthraugh the slndgle slit caherentlwe can avaiathis Step to ay usan lasers sienna up Mawoi nilv m plluvI mm o the light from the sin e slit is then used to illuminate a doubles It which produces an Interference pattern on a screen behln IL Knuth t syn mammmmmi WWW o BTW note the interference pattern due to the WM double slit 7 it looks like a sinewave 7 What lsthe FaurlerTransfarm of a quotdouble Slit thmhmlmrlum ammonium Puntz Null lnl ZDDBnateschzs 28 ah SlEal DDtlES aat Ultrasonic Imaging o Ultrasonic Imaging involves the following Generation of acoustic wavelets Control of the timing and amplitude of these wavelets oThe goal is to control regions of constructive and destructive interference in other words to form beams Reception and processing of the echoes to form the image Overall Block Diagram of an Ultrasound Scanner Transmit Beamronnation Transducer Array Image Receive Formation Beamronnation 0 Beamformation generation of coordinated timing signals for transmit and delays for receive processes Probably the most expensive building block Acoustic Wave Propagation 0 Transducers usually multi element arrays or piezoceramic elements 0 Image formation conversion to video raster image processing Ultrasound Imaging Eiectrodes Transduction tz re iBeamformation Cause Stress Sequence of Events in Scanning 0 System processor initiates a scan 7 former set to appropriate Scan 0 ie I Transmitiocai iocation e Puising voitage appiied to array eiements in timed sequence 7 Sound emitted from eiements e Prerampiitiers initiate reception of ecnoes 7 Receive beamtormer appiies needed deiays to optimize energy from desired focus and ioo angie 7 Data stored scan conversion to video raster egins 0 Process repeats itself Ultrasound Probes Convex Linear IntraoperatiVe Vaginal Rectal Ultrasound Tronsducer Structure Element Isolation Cuts Acoustic Lens Acoustic Matching Layers Ground Electrode Piezocerarnic Signal Electrode Acoustic Absorber Ground Flex J Signal Flex Circuit Scan Control Bus in Front End Control Bus Video Timing System Controi Bus yME Bus Beamformation Focusing Basic focusing type beamformation Symmetrical delays about phase center Beamformation Focusing and beam steering Beam steering W linear phased arrays Asymmetrical delays long delay lines Geometry for Beamsteering Calculations 0 Delay determination simple path length difference reference point phase center apply Law of Cosines approximate for ASIC implementation 0 In some cases split delay into 2 parts beam steering dynamic focusing Linear Scan Transmit Beamforming u 1 2 g nmnpmam n Focused Beam Water path Step n Tima mm us WeidlinqerAssociates Flex version 1 I15 MH H MHHHHJA WM u H39HH HN H MH W MHth m materials de 50MHZ 1 D linear array beam focused m 5 cm Still images from sequence 5mm mummy mmmmm swm pomuuruummmsm Scon Conversion nun u m u m u m u u u u Vectors om acoustic space must be scan convened to display pixels rasterized nommal 400 x 400pxxels 7 R o to XX 7 XX to XX Ultrasound Scanners Most often prescribed modality ilnstmment sales 4B General Imaging OBGyn 77 Cardiac Variety of Applications g 39 u gt Portable L i r gt LowCost Cariac Kidney Liver gt Safe gt RealTime Obstetrics Vascular Scanning Modes BScan Real Time All Mode There are several additional modes mainly involving Doppler These will be covered in later lectures Another Propagation Model oSpatial Impulse Response 39 Method Assume a transducer generates a delta function motion The temporal response of this can be determined for all field points 0 This is the spatial impulse response Once this is known a field can be determined for an transducer driving function J M 3 La mlnisantelmn m39 Tess What is the spatial impulse responSe at the focus of a circular aperture Field II coherent field simulation httpwwwesoersted sssreees r seems er srsrseeeer m 0 dtudkstaff a field ZTZZJSEEJ Semi Z 5miir sesr elements m Matlab simulation of 39 2316 an ultrasound scanner But can be readily Transducer center frequenCy Hz Sampllng frequency Hz a Deflne the transducer Th xdciconcave R Rfocus eleislze converted to any coherent probe hee1nPre response ere erersssres er impuleeireeponeeeeinzepirror o 1fsz2f0 Homework rm erse res essesrm s assignments to be slieumiulsisreepen e e Xdcilmgulse Th 1mpulseiresponse based on Field excltatlon51n 2p1ro o s 1 Xdciexcltatlon Thy 7 lseirespon ee hann 1ng max 39 o o 5e2f0 excitatlon Trdnsducer to be simulated Single circular element 8 mm radius Analysis grid 1 mm Center frequency 3 MHz Field39s internal sampling frequency 100 MHz r mm y mi xdcizmpulse Th Field Code Review x get anataai parameters Reemoo x Radius of transducer mi Rfocus 01000 r Geo etrac focus pcant mi eleislzezlJOOO x size of mathematicai eiements mi f03e5 Transducer center frequency Hz Hz x t fszlooea x Sampiang frequency x Define the transducer Th xdciconcave R Rfocusk eleiszze x Se tne ampuise respcnse and excitaticn of the emat aperture nipuseiresponsezslne p 0 01fs2f0 pcnse iaannangltmaxlt gtgt nipulseixesponse excitationzsan FI EO Ofs2f0 xdciexcltatlon Th excatatacn I Call to xdciconcave 7 Purpose Procedure for creating a concave transducer Culling Th xdc concave radius i ocal radius ele size 7 Input Radius Radius of physical emerits focal radius Focal radius ele size Size of mathematical elements 7 Output Th A pointerto this ucer aperture De ne transducer impulse response 7 Hannin eweighted sine wave I De ne the transmit pulse 7 Excitation e sine wave nipulseiresponse nipulseiresp Field Code Review x Define the transducer Th e xdciccncave R Rfocus eleislze x get the lmpulse respcnse and excatatacn of the emat aperture inlt2 pi uise cm 01fs2f0 respcnsexnannrngmuu gt xdcimipulse Th impulseiresponse o Size nipulseixesponsen excatatacnesanlt2 pa to a Jfsc2f0 gt xdciexcltatlon Th excitation x eaicuiate tne puise ecnc field and dispiay n t xpolntsr00210 RFidatx startitime e calcihhp Th Th xpolnts zeros0 ZO ones101 000 0 Call to calchhp Field II Code Review Purpose Procedure for calculating the pulse echo field Calling hhp starttime calc hhplTh1Th2 points Input 0 Th1 Pointerto the transmit aperture 0 Th2 Pointerto the receive a erture 0 points Fleld points Vectorvvlth t ree columnslxvz and one row for each field point Output hhp Received voltage trace start time The time for the first sample in hhp calculate the pulse echo fleld and dlsplay 1t xpojntseloo2lo RFidatm startitlm xpolnts zerosl101i c cihhp Th Th 300nes1101 1000i Make a dlsplay of the envelope Jgureu envabshllbertRFidata15600 m env204rloglo armmax max em NMslzeenv envenv6oenvgteso a 60 meshxpolnts 0Nelfs startitlme rle6 en ylabel Tlme mus xlabel Lateral distance m i tltle Pulserecho fleld from 8 mm Concave transducer at 30 mm 10 10 3341 396 60 0 vlew14 80 axls Selected Fie oGeneral Field ll Commands 0 Function name Purpose 7 field debug lnitlalize debugglng field end Terminate the Field Id Commands Page ll program svstem and release the storage a field init lnitlalize the Field ll program svstem 13 a set sampling Set the sampling frequencv the svstem uses 14 a set field Setvarious parameters for the program 14 OTransducer or array commands a gtltdc apodization Create an apodization time linefor an aperture 16 e gtltdc baffle Set the baffle condition for the aperture 16 e gtltdc center focus Set the origin for the dvnamic focusing line 17 a gtltdc concave ine oncave aperture 17 a gtltdc convert Convert rectangular description to triangular description 18 a xdc convex arrav 18 Create a convex arrav transducer 19 transduceri Llulluu l multirrow 20 Selected Field Commands ITrunsducer or array commands a m e dynuml fucus Use dynuml fucuslng furun aperture 22 e m exettattah Setlhe Excltutlun pulse ufun aperture 23 e raeraeus createaraeus tlmE llnefurun a Ermre p e xdcfucused may create an Elevutlunfucused llneururmy transaueer 23 a m fucused mulnruw Create an ElEvutan fucused llneur multH DW transducer 2H 7 raefree perture 27 Free starage uncupled by arr a e m get Getlnfurmutlun apautarr upermre 27 e m lmpulse Setthe lmpulserespunse uf an aperture 29 e KdE lln create a llneururmy transducer 3n 7 m llneurmultlmw create a hrrearrnultrraw army transaueer 3n IHeld COlCUlOUOn Commands 7 ul Calculate mesputlul lmpulserespunsa ua e ealehhp Calculate the pulse Echufleldt Sn 7 ul hp Calculate the ermttea eld 51 7 ul seat Calculate the renewed slgnulfrumu cullectlun uf scatterers 52 Pulse echo Field from O Concave Cm Irma Pusan eld ram 3 mm mum lunsdttcay at an m m lnlzml mums mm Homework H Download Field II and the User39s Guide to your computer Execute the first example from httpwwwesoersteddtudkstaff39a39fieldexampleshtml the point spread function logo case Change the number of transmitted cycles to one cycle three cycles How does the point spread function change Change the field location to 80 mm from 30 by editing the following statement RFidata startitime calcihhp Th Th Xpoints zeros1101 30ones1101391000 Describe the likely impact on quality of imaging of the two point spread functions Hand in your graphical matlab results N 94 P 0quot Summary 0We have reviewed the major building blocks of an ultrasound scanner Transduction Beamformation Scan conversion 0We have been introduced to the Field simulation program and walked through some of the code Instructor Contact Information Badri Roysam Professor of EIectrIcaI Computer amp Systems Engineering Office JEC 7010 Rensseiaer Poiytecnnic Institute 11018m Street Troy New York 12180 Phone 518 27678067 Fax 518 27662612433 ma39l ro sam ecse e Website hug www rgi eduNroysab NetMeeting ID for offcampus studenis 128 113 61 80 Secretary Laraine MichaeiidesJEC 7012 518 276 78525 michairgi edu Renssolaer Instructor Contact Information Kai E Thomenius Chief Technoiogist UItrasound amp Biomedicai Office KW7C300A GE Giobai Research Imaging Technoiogies Niskayuna New Vork 12309 Phone 518 38777233 Fax 518 38776170 Email thameniu crd ge com nomeniusecsergi edu Secretary Laraine Michaeiides JEC 7012 518 276 78525 micnairgi edu Rensseiaer GE Global Research BMED 4962ECSE4962 Introduction to Subsurface Imaging Systems Lecture 3 X ray Imaging Kai E Thomenius1 amp Badri Roysam2 1Chief Technologist Imaging Technologies General Electric Global Research Center ZProfessor Rensselaer Polytechnic Institute GE Global Research Center for Subsurface magmg amp Sensmg Review of Last Lecture 0 Methods for classifying different imaging probes 7 Coherence 7 Interaction with target 7 Projection or pulseecho Xrays as an example for those classi ers Today 7 Projection Imaging 7 Xrays Outline of Course Topics PULSE ECHO METHODS THE BIG PICTURE What is subsurface imaging Examples Why a course on this topic MRI EXAMPLE Projection Imaging Adifferent sensing modality XRay Waging 39om the othe s Basics ofMRI Computer Tomography COMMON FUNDAMENTALS Propagation ofwaves Interaction of waves with targets of interest MOLECULAR IMAGING What is it PET amp Radionuclide Imaging IMAGE PROCESSING amp CAD History Nov 1885 Rontgen discovers X rays in Wurzburg German 7 Observations of uorescence when using a Crookes tube Initial results published in Dec 28 1885 broader announcements in January 1 86 In the next several months reports of diagnoses of fractures bullets even uoroscopy started coming Late in 1886 Siemens amp GE started selling Xray equipment IIIIHIIIIS inn PELYIX Electromagnetic radiation photons Xray Generation Bremsstrahlung Electrons Xrs 5 Low Energy A Low Frequency Higi Energy B High Frequency Wavelengths 7i ranges from 10 pm to 10 nm Propagation speed cn 300000 krnsec Un ltered Brems Energy of xray photon Inten imy Characteristic 1w where h Planck s constant or i 6 e s v frequencyc Usefulrelation A o eq v Quick Example What is the frequency and energy of those Xrays in a beam Whose wavelength A 1 nm Answer 0 re v f q A Ehvl99eil Jl24KeV 3617 Hz 300000 THz Xray Spectra Un ltered Brems Electron volt Inten 7 1 eV is the energy increase that an sity electron experiences when accelerated over a potential difference of l V 7Or 1 eV qg AV 16e19J Most of Xray applications in medical Characteristic 7 xrays imaging are in range 0 Photon energy 720 KeV lt E lt 150 KeV These energies correspond to frequencies Filtering With high Z mate al in the range of absorbs low energy Xrays and 748e18 Hz lt v lt 36el9 Hz shapes the output spectrum to And wavelengths in the range of reduce dose to patent 7625e9 cm lt A lt 833e10 cm Characteristic Xrays xray Lelectron In addition to brems Xrays out electron a tube usually generates from 77 characteristrc Xrays lament knocker Steps electron 7 Fast moving electron knocks off an inner orbital electron 7 Vacancy created is lled by a higher orbital electron 39 v I t f 7 The energy thus released is l mpac 9 highly quantized and shows up acceleranon as a sharp spectral peak voltage Xray Interactions With Matter Primary effects at medical energies 7 Photoelectric effect Energy of xray is absorbed by an orbital electron Electron is ejected from atom 7 Scatteri Compton scatter 7 Some of the xray energy transferred to e an el ctron Xray travels on 2 3 with an altered direction and less energy 7 Coherent scatter 7All ofxray l energy interacts with the atom quot but is later reradiated with l same energy in an arbitrary k7 n 71 1 i direction Antiscatter grid Schematic of 3 Diagnostic Xray System Higm anmge Cable Transform er 250 v tn 1m kV Control Unit Mammography Tubes Film Cassette Scintillator Film XRAY SCINTI LLATOR FILM SCINTI LLATOR Enclosure Light tight Xray transparent FILM CASSETTES Xray generation is a fairly in ac Photon energy is related to Total number of photons is Xray Ef ciencies PVI l ef cient process Most of the electrical power ends up as heat at the anode Of the generated Xrays only 2 end up being absorbed at the detector celeration voltage eg 10 KV will produce up to 10 KeV photons related to cathode current Typical currents are in few mA Dangers of X rays Xrays are ionizing radiation will cause tissue arnage 1rad 100 ergs of energy gram oftissue Frame of reference 7 Typical chest xray exposure 10 100 Inmds 7 Background mdiation in upstate NY 200 7 300 mmds 7 Exposure of50 mds causes mdiation sickness 7 Dosages exceeding 200 mds can cause death 7 Geneml public should absorb less than 05 radsy 7 Radiation than less ByJune 18 dangers amount of Xray radiation that imparts ear workers less than 01 radweek and 5 mdsyear 86 workers were cautioned of the associated w Xrays Medics XRays 11 Casmic 9 Internal 11 XRay Detectors TODAY Phosphor Film r amp XRay XRay gm a MA GE FL UOROSCOPY TODA Y Image lntensi er X39R Y may Light ANALOG J IMAGE TV Pickup Electron Tube ALL APPLICATIONS DIGITAL DETECTOR etector DIGITAL IMAGE Pos 1t10ners 39Positioners determine how an x ray system looks So how is an image actually formed ELE Filte quot Collimator Patient Antiscatter grid Scintillator A Detector Cranial XRay Mammography 0 Screening diagnostic and interventional procedures for breast imaging 0 Only true Xray screening modality Image Review Senographe 2000D Workstation Rad Rad is most widely used Xray procedure Revolution XRd RampF Classical and remote rampf systems GI is major application barium is contrast agent Single Contrast GI Exam Precision 500D Surgical Minimally invasive interventional and intraoperative procedures Orthopedics general surgery vascular OEC Series 9800 Vascular Interventional cardiac and peripheral vascular imaging Fluoro widespread uoro IQ critical Most exams use iodine contrast agents Innova 4100 LambertBeer Law 2 a 6 where 7 I z is the xray intensity at the measurement plane 7 In is the xray intensity at the source plane 7 z is the distance between the source amp msmt planes 7 u is the attenuation coefficient 0 Also known as Beer Lambert or Beer s Law 39 Intensity can be given as photon density photonmAZ dx gt d Linear attenuauon coef cientquot l g 1quot x 1 l gt ID Beer39s lawquot x Iex17IIIII H In general u is a function of x this complicates matters Inhomogeneltle s amp LambertBeer Consider a multilayer object to be examined LambertBeer Law has to be applied to each segment With a constant p In the extreme With a continuously varying p the application of this law has to be as a line integral 12 1expyl 11 l l 1 Inexpyl 15 LgtEXP39lA IIJ Xray contrast De nition for present 7 Difference in the x ray IO 10 intensities emerging from adjacent regions Sometimes given as a 12 1 111 percentage 1 III Film can17asl 2 1gtltlOO 2 or as a normalized percentage contrast 2 J11 X l 00 2 l Xray contrast factors Contrast is caused by differential attenuation in the independent paths Here are some factors that have an impact on the amount of contrast observed 7 Thickness of the subject 7 Effective density of tissue along the path 7 Energy of the Xray beam KV Higher KV higher energy lower contrast Lower KV lower energy higher contrast Our interest is in the 2nd item tissue density The other parameters should be set so as to maximize the observed contrast Contrast increasing complexity Consider cases with different thicknesses 1 1 andor attenuations 11 Ige l lxl 12 Ige l lxl l l log1 loglt7 ulx1 gtlt loge x1 H1 log2 loglt7 L12x2 gtlt loge 2 x2 logfz 10g11 we 5le X loge Ill 12 l The ch product is critical This product establishes detectability criterion for Xray systems Contrast and Signal to Noise Ratio Signal to noise ratio SNR gives a measure ofhow far a signal is above the noise in a sys em Quantum and electronic noise are inherent in any imaging chain Their minimization is critical for the system designer Diagnostic potential is related to how far critical contrast features are above the noise ContrasttoNoise Ratio Quantum noise For a digital xray detector system with square pixels 7 if the average number of xrays recorded in each pixel is N 7 then the noise per pixel will be 04 Statistical distribution associated wu whomquot 10cc norm Average par mu Amiga par mu 0 r ma mo 9 Handnm Pnuem enn RncLnmr with xrays is the distribution 7 e above relation falls out directly from this fact Standard mmquot sranaaru nemtian m airman mat has nhntonIJ 3 Signaltonoise ratio The signaltonoise ratio SNR is given by N N SNR 7 g 7 W 7 W When the number of xrays N is increased the radiation dose also increases To double the SNR the dose to the patient needs to be increased by a factor of 4 Contrasttonoise ratio CNR for any two intensities 11 and 12 at a detector is given by CNRL7IZIL7IZI 039 N 7 Here N is the nominal value of photons reaching the detector Receiver Operating Characteristic ROC curves Most basic task of the diagnostician is to separate abnormal subjects from normal subjects In many cases there is significant overlap in terms of the appearance of the image 7 Some abnormal patients have normallooking lms 7 Some normal patients have abnormallooking lms ROC curves are a tool for assessing the performance of a hypothesis testing algorithms 2 X 2 Decision Matrix Actually Abnormal Actually Normal Diagnosed as True Positive False Positive Abnormal TP FP Diagnosed as False True Negative Normal Negative FN TN ROC curves cont For a single threshold value and the population being studied a single value for TP TN FF and FN can be computed The sum TP TN FP FN will be equal to the total number of normals and abnormals in the study population True diagnosis must be determined independently based on biopsy con rmation longterm patient follow up etc ROC Curve declslon threshold true positive fractwon of patients o 2 t1 decision cmtericn false positive fraction B Summary Quick historical review of X rays was given Block diagrams key components de ned Brief discussion of X ray scattering 39 An X ray beam traversing through an object is attenuated by the exponential Lambert Beer Law The product of the attenuation coef cient and the path length of the X ray beam in such a target is critical in establishing detectability Next tiIne X Ray Metrics Homework 6 Cm 4 D 1 1 Breast Detector Given a rectangular slab of breast tissue 6 cm 39c Assume there is a lesion size ofl cm embedded in the breast At 20 KeV the linear attenuation coef cient ofglandular tissue pg 08 and that ofthe lesion pl 09 Assume that there are 10 1000 photons incident at the breast By determining the CNR for the lesion comment on its detectability What ifthe lesion is 05 cm Instructor Contact Information uyszm Professor ofElectncal Computer lt1 Systems Engmeenng Office EC 7010 Rensselaa Polytechnic instmne roy New York 12180 Phnne 518 27678067 Fax 518 27 6762612433 Email roysamecse1 edu Websim mt www reduNro sab NetMeeLing 1 far utpcampus smdmu 128 113 61 80 Secretary Betty Lawson DEC 7012 518 276 785251awsob m edu Rem eh6r Instructor Contact Information Kai E Thomenius ChiefTeclmologist Ultrasound amp Biomedical Of ce KWC300A GE Global Research Imaging Technologies Niskayuna New York 12309 Phone 518 3877233 Fax 518 3876170 Email Lhomeniu crdgecom Lhomenius ecse iedu Secretary TBD Ren elaer V GE Global Research