Bioengineering 508: Physical Aspects Of Medical Imaging

Bioengineering 508:Physical Aspects of Medical ImagingBioengineering 508:Physical Aspects of Medical oduction to Medical Imaging1. Medical Imaging Modalities2. Modern Image Generation3. Intro to Image QualityOrganizer: Paul Kinahan, PhDAdam Alessio, PhDRuth Schmitz, PhDLawrence MacDonald, PhDAdam Alessio, PhDDepartment of RadiologyUniversity of Washington Medical Centeraalessio@u.washington.eduImaging Research epartment of RadiologyUniversity of Washington Medical CenterAlessio - BIO508Nature of Medical ImagingAlessio - BIO508Nature of Medical ImagingFor this class:Medical Imaging: Non-invasive imaging of internalorgans, tissues, bones, etc.Focus on:1. Macroscopic not microscopic2. in vivo (in the body) not in vitro (“in glass”, in the lab)3. Primarily human studies4. Primarily clinical diagnostic applicationsAlessio - BIO508QUICK CAVEAT Powerpoint Slides are just a vehicle for major topicsThese do not have all the information discussed inclass!Taking notes to supplement slides is probably agood idea!Alessio - BIO5081

Types of Medical Imaging (Modalities)Types of Medical Imaging (Modalities)Electromagnetic SpectrumGrouped by underlying physics: X-Ray/CTMajor 4 that dominate Ultrasoundclinical imaging, focus Magnetic Resonance Imaging (MRI)of this course Nuclear Medicine OpticalPrimarily microscopic Magnetic Field Electric FieldMainly research based Thermal Optoacoustic ElastographyAlessio - BIO508Alessio - BIO508Types of Medical Imaging (Modalities)Classifications of Medical Images1. Anatomical vs. Functional Nuclear medicineModern Image GenerationFrom continuous real world to a meaningful image(on computer):1. Sampling Continuous InformationAnatomy/Structure/Features vs. Physiology– Information and sampling technique varies widely for eachmodality- Topic for later lectures– Computer can only hold discrete chunks of data– Pixel a single picture element; Voxel a single volumeelement2. Emission vs. Transmission Where does energy imaged originate?3. Projection vs. Tomographic Alessio - BIO508Projection-- 2D imaging, single plane, no depthinformationTomographic (“tomo” slice, graphy image) -- volumetricFor comparison, this iswavelength/frequency range of US,but US is NOT electromagnetic!2. Quantizing Samples– Each discrete chunk must be represented by certain numberof bits3. Visualization Techniques of quantized, sampled imagevolumesAlessio - BIO5082

1. Sampling Continuous InformationGiven a signal such as a sine wave withfrequency 1 Hz:Alessio - BIO508Intro to Sampling TheoryWe can also sample the signal at a slower rate of2 Hz and still accurately reconstruct the signal:Alessio - BIO508Intro to Sampling TheoryWe can sample the points at a uniform rate of 3Hz and reconstruct the signal:Alessio - BIO508Intro to Sampling TheoryHowever, if we sample below 2 Hz, we don’t haveenough information to reconstruct the signal, and infact we may construct a different signal (an alias):Alessio - BIO5083

Intro to Sampling Theory Intro to Sampling TheoryAliasing– occurs when your sampling rate is not high enough to capture theamount of detail in your image– Can give you the wrong signal/image—an alias– Where can it happen in graphics? During image synthesis:– sampling continuous signal into discrete signal– e.g. ray tracing, line drawing, function plotting, etc. To perform sampling correctly in image space, needto understand structure of data/image During image processing:– resampling discrete signal at a different rate– e.g. Image warping, zooming in, zooming out, etc. Fourier: “Any periodic function can be rewritten as a weightedsum of sines and cosines of different frequencies.” - FourierSeriesNyquist criterion: Must sample at two times the highest frequency in thesignal for the samples to uniquely define the given signalFNyquist SamplingRate2– Sampling below the Nyquist frequency can cause aliasing (CD sampling example)Alessio - BIO508Alessio - BIO508A sum of sines Our building block: Add enough of them to getany signal f(x) you wantWhich one encodes thecoarse vs. fine structure ofthe signal?What would an image looklike with a lot of highfrequency content?What could you do to reducespeckled noise from animage? Asin("x ! )Fourier TransformSignal f(x)1D Example: A signal composed of two sinewaves with frequency 2 Hz and 50Hz The Fourier Transform of thesignal shows these twofrequenciesIn 2D: Usually represent lowfrequencies near origin, highfrequencies away from originHigh FreqHigh FreqfrequencyLow FreqHigh FreqAlessio - BIO508Fourier Transform of f(x)High FreqAlessio - BIO5084

2D Fourier TransformsImage in frequency domainImage in space domain (magnitude of frequency component)2D Fourier TransformsImage in frequency domain(log magnitude of frequency component)Image in space domainImage in frequency domain(magnitude of frequency component)Image in frequency domain(log magnitude of frequency component)OriginalAfter low-passAfter high-passAlessio - BIO508Frequency ContentAlessio - BIO508Alessio - BIO508Frequency ContentAlessio - BIO5085

Modern Image Generation2. QuantizationFrom continuous real world to a meaningful image(on computer):1. Sampling Continuous Information– Information and sampling technique varies widely for eachmodality- Topic for later lectures– Computer can only hold discrete chunks of data– Pixel a single picture element; Voxel a single volumeelement2. Quantizing Samples– Each discrete chunk must be represented by certain numberof bits Only have finite storage available for each pictureelement Digital images have “digitized” intensity values.Continuous values are quantized into discrete values.– Example: “Truecolor” on computer displays use 24 bits foreach pixel (8bits blue, 8 bits red, 8bits green 256x256x256possible colors)– Many medical imaging modalities use intensity values of 12bits per pixel. (2 12 4096 possible gray levels)3. Visualization Techniques of quantized, sampled imagevolumesAlessio - BIO508Alessio - BIO508Color depth8 bits per pixel5 bits per pixel4 bits per pixel3 bits per pixel2 bits per pixel1 bit per pixelAlessio - BIO5086

Nature of Medical Imaging For this class: Medical Imaging: Non-invasive imaging of internal organs, tissues, bones, etc. Focus on: 1.Macroscopic not microscopic 2.in vivo (in the body) not in vitro ("in glass", in the lab) 3.Primarily human studies 4.Primarily clinical diagnostic applications Alessio - BIO508 Nature of Medical Imaging QUICK .