digital radiography – chapter 11 adjuncts to radiology – chapter 12 brent k. stewart, phd, dabmp...

Digital Radiography – Chapter 11Digital Radiography – Chapter 11Adjuncts to Radiology – Chapter 12Adjuncts to Radiology – Chapter 12

Brent K. Stewart, PhD, DABMPBrent K. Stewart, PhD, DABMPLois Rutz, M.S. Radiation Safety Engineering, Inc.Lois Rutz, M.S. Radiation Safety Engineering, Inc.

a copy of Brent Stewart’s unmodified lecture may be found a copy of Brent Stewart’s unmodified lecture may be found at:at:

http://courses.washington.edu/radxphys/PhysicsCourse04-05.htmlhttp://courses.washington.edu/radxphys/PhysicsCourse04-05.html

Take Away: Five Things You should be able Take Away: Five Things You should be able to Explain after the DR/Adjuncts Lectureto Explain after the DR/Adjuncts Lecture

The various types of detectors used in digital imaging The various types of detectors used in digital imaging (e.g., scintillators, photoconductors, etc.)(e.g., scintillators, photoconductors, etc.)

The differences between the various technologies used The differences between the various technologies used for digital radiography (e.g., CR, indirect and direct DR) for digital radiography (e.g., CR, indirect and direct DR)

Benefits of each type (e.g., resolution, dose efficiency)Benefits of each type (e.g., resolution, dose efficiency) Why digital image correction and processing are Why digital image correction and processing are

necessary or useful and how they are executednecessary or useful and how they are executed The various types of adjuncts to radiology (e.g., DSA or The various types of adjuncts to radiology (e.g., DSA or

dual-energy imaging), what issue they are trying to dual-energy imaging), what issue they are trying to resolve, mechanism exploited and end resultresolve, mechanism exploited and end result

Why Digital/Computed Radiography Why Digital/Computed Radiography

Limitations on Film/Screen radiographyLimitations on Film/Screen radiography

Screen/Film system is image receptor and displayScreen/Film system is image receptor and display Image characteristics depend on Screen/Film and Film Image characteristics depend on Screen/Film and Film

processing.processing. Modification of image difficult to control (e.g. development Modification of image difficult to control (e.g. development

temperature).temperature). Image appearance depends on technique settings.Image appearance depends on technique settings. Image quality cannot be repaired after development. Retake Image quality cannot be repaired after development. Retake

only solution to poor I.Q.only solution to poor I.Q.

Why Digital/Computed Radiography cont.Why Digital/Computed Radiography cont.

Screen/film dynamic range 2 to 2.5 orders of magnitude. Screen/film dynamic range 2 to 2.5 orders of magnitude. Different applications require different screen/film combinations. Different applications require different screen/film combinations.

Only one “original” image.Only one “original” image. Films often “go missing” from ER or ICU and never are archived.Films often “go missing” from ER or ICU and never are archived. Copies expensive, have inconsistent quality, and often are non-Copies expensive, have inconsistent quality, and often are non-

diagnostic.diagnostic. Archive space expensive, often remote.Archive space expensive, often remote. Digitizing film is only way to move images to PACS.Digitizing film is only way to move images to PACS.

How does Digital/Computed Radiography solve How does Digital/Computed Radiography solve these problems?these problems?

• Decouples imaging chain components.

• Detector, image processing, display all “independent” entities.• Independent in design but not in application.

• Detector can make use of extended dynamic range.• Solid state detectors have improved DQE.

• Electronics can apply corrections to input signals.• In particular, over/under exposure can be corrected, reducing

retakes.

How does Digital/Computed Radiography solve How does Digital/Computed Radiography solve these problems? Cont.these problems? Cont.

• Image processing can modify and enhance raw (pre-processed) data.

• Images can be displayed on workstations which permit interactive display processing.

• Image data is stored digitally. “Original image” is available everywhere and at any time.

CR vs. DRCR vs. DR

CR also known as a Photostimulable Phosphor system.CR also known as a Photostimulable Phosphor system. CR uses an imaging plate similar to an intensifying screen as the CR uses an imaging plate similar to an intensifying screen as the

receptor.receptor. CR systems are indirect digital systems.CR systems are indirect digital systems.

Indirect systems convert x-radiation to the final digital image through Indirect systems convert x-radiation to the final digital image through one or more stages. one or more stages.

DR digital radiography DR digital radiography Uses a fixed detector such as amorphous selenium plate as the Uses a fixed detector such as amorphous selenium plate as the

receptor.receptor. Can be a direct or an indirect digital system.Can be a direct or an indirect digital system.

When direct it is sometimes called DDR for direct digital radiographyWhen direct it is sometimes called DDR for direct digital radiography

CRCR

Detector or Imaging Plate (IP) is essentially a type of Detector or Imaging Plate (IP) is essentially a type of intensifying screen.intensifying screen. IP can be used in any bucky or table-top system.IP can be used in any bucky or table-top system. IP is relatively robust. Requires same care as intensifying IP is relatively robust. Requires same care as intensifying

screens.screens.

Process is indirect. Process is indirect. X-ray creates excitation center. X-ray creates excitation center. Plate reader uses red light to stimulate centers to release blue Plate reader uses red light to stimulate centers to release blue

light.light. Blue light is directed to a photo-electric transducer (pmt or Blue light is directed to a photo-electric transducer (pmt or

other).other). Electric signal digitized to make raw image.Electric signal digitized to make raw image.

CR and DR SystemsCR and DR Systems

Image Production in CR/DR SystemsImage Production in CR/DR Systems

Radiation through the patient creates a latent image on the receptor.Radiation through the patient creates a latent image on the receptor. Receptor is “read” by some process and latent image is converted Receptor is “read” by some process and latent image is converted

to an electronic signal.to an electronic signal. Signal is processed.Signal is processed.

Processing is related to acquisition system characteristics.Processing is related to acquisition system characteristics. Signal (analog) is converted via ADC to a bit value in a digital Signal (analog) is converted via ADC to a bit value in a digital

matrix.matrix. Digital image is processed.Digital image is processed.

Processing is related to desired image information.Processing is related to desired image information. Digital matrix is displayed on a video screen or printed to paper or Digital matrix is displayed on a video screen or printed to paper or

film.film.

Signal ProcessingSignal Processing

Primarily to accommodate variations in the Primarily to accommodate variations in the detector/electronics components.detector/electronics components.

Involves corrections for dead space, non-uniformities, Involves corrections for dead space, non-uniformities, defects.defects. Could be developed to compensate for MTF losses.Could be developed to compensate for MTF losses.

All systems, PSP or Direct, do some sort of processing All systems, PSP or Direct, do some sort of processing and scaling.and scaling.

Ultimate goal is to present the image processing module Ultimate goal is to present the image processing module with “true” image pixels.with “true” image pixels.

Digital Image CorrectionDigital Image Correction

Interpolation to fill in dead pixel and row/column defectsInterpolation to fill in dead pixel and row/column defects Subtracting out average dark noise image DSubtracting out average dark noise image Davg(t)avg(t)(x,y)(x,y) Differences in detector element digital values for flat fieldDifferences in detector element digital values for flat field

Gain image: G(x,y) =G’(x,y) - DGain image: G(x,y) =G’(x,y) - Davg(t)avg(t)(x,y); G(x,y); Gavg avg =(1/N) ∙ =(1/N) ∙ G(x,y) G(x,y)

Make corrections for each detector element (map)Make corrections for each detector element (map) I(x,y) = GI(x,y) = Gavgavg ∙ [I ∙ [Irawraw(x,y) - D(x,y) - Davg(t)avg(t)(x,y)] / G(x,y)(x,y)] / G(x,y)

Done for DR and in a similar manner for CT (later)Done for DR and in a similar manner for CT (later) Not performed for CR on a pixel by pixel basis, although Not performed for CR on a pixel by pixel basis, although

there are corrections on a column basis for differences in there are corrections on a column basis for differences in light conduction efficiency in the light guide to the PMTlight conduction efficiency in the light guide to the PMT

Digital Image CorrectionDigital Image Correction

c.f. Bushberg, et al. The Essential Physics of Medical Imaging, 2c.f. Bushberg, et al. The Essential Physics of Medical Imaging, 2ndnd ed., p. 310. ed., p. 310.

DetectorsDetectors

In order to understand signal processing we need to In order to understand signal processing we need to learn about the detectors.learn about the detectors. Photo Stimulable Phosphor PlatesPhoto Stimulable Phosphor Plates Photoconductive materials.Photoconductive materials.

Detector consists of a receptor material (e.g. BaF(H)Eu), Detector consists of a receptor material (e.g. BaF(H)Eu), and a set of signal readout and conversion electronics.and a set of signal readout and conversion electronics.

Receptor responsible for the DQE.Receptor responsible for the DQE. Rest of the system contributes to noise, resolution,dynamic Rest of the system contributes to noise, resolution,dynamic

range.range.

Detectors in Digital Imaging (1)Detectors in Digital Imaging (1)

Gas and solid-state detectorsGas and solid-state detectors Energy deposited to e- through Energy deposited to e- through

Compton and photoelectric Compton and photoelectric interactionsinteractions

Gas detectors – apply high Gas detectors – apply high voltage across a chamber and voltage across a chamber and measuring the flow of e- measuring the flow of e- produced by ionization in the produced by ionization in the gas (typically high Z gases like gas (typically high Z gases like Xenon: Z=54, K-edge = 35 Xenon: Z=54, K-edge = 35 keV)keV)

Were used in older CT unitsWere used in older CT units

c.f. Bushberg, et al. The Essential Physics of c.f. Bushberg, et al. The Essential Physics of Medical Imaging, 2Medical Imaging, 2ndnd ed., p.32. ed., p.32.


Solid-state materialsSolid-state materials Electrons arranged in bands with conduction band usually emptyElectrons arranged in bands with conduction band usually empty

Solid-state detectorsSolid-state detectors Scintillators – some deposited energy converted to visible lightScintillators – some deposited energy converted to visible light Photoconductors – charge collected and measured directlyPhotoconductors – charge collected and measured directly Photostimulable phosphors – energy stored in electron trapsPhotostimulable phosphors – energy stored in electron traps

c.f. Yaffe MJ and Rowlands JA. Phys. Med. Biol. 42 (1997), p. Elements of Digital Radiology, p. 10.c.f. Yaffe MJ and Rowlands JA. Phys. Med. Biol. 42 (1997), p. Elements of Digital Radiology, p. 10.


c.f. Yaffe MJ and Rowlands JA. Phys. Med. Biol. 42 (1997), p. Elements of Digital Radiology, p. 9.c.f. Yaffe MJ and Rowlands JA. Phys. Med. Biol. 42 (1997), p. Elements of Digital Radiology, p. 9.

Computed Radiography (CR)Computed Radiography (CR)

Photostimulable phosphor (PSP)Photostimulable phosphor (PSP) Barium fluorohalide: 85% Barium fluorohalide: 85%

BaFBr:Eu + 15% BaFI:EuBaFBr:Eu + 15% BaFI:Eu ee-- from Eu from Eu2+2+ liberated through liberated through

absorption of x-rays by PSPabsorption of x-rays by PSP Liberated eLiberated e-- fall from the fall from the

conduction band into ‘trapping conduction band into ‘trapping sites’ near F-centerssites’ near F-centers

By low energy laser light (700 nm) By low energy laser light (700 nm) stimulation the estimulation the e-- are re-promoted are re-promoted into the conduction band where into the conduction band where some recombine with the Eusome recombine with the Eu3+3+ ions ions and emit a blue-green (400-500 and emit a blue-green (400-500 nm) visible light (VL)nm) visible light (VL)

c.f. Bushberg, et al. The Essential Physics of Medical c.f. Bushberg, et al. The Essential Physics of Medical Imaging, 2Imaging, 2ndnd ed., p. 295. ed., p. 295.

Computed Radiography (CR) System (1)Computed Radiography (CR) System (1)

Imaging plate (IP) made of PSP is Imaging plate (IP) made of PSP is exposed identically to SF exposed identically to SF radiography in Buckyradiography in Bucky

IP in CR cassette taken to CR IP in CR cassette taken to CR reader where the IP is separated reader where the IP is separated from cassettefrom cassette

IP is transferred across a stage IP is transferred across a stage with stepping motors and scanned with stepping motors and scanned by a laser beam (by a laser beam (~~700 nm) swept 700 nm) swept across the IP by a rotating across the IP by a rotating polygonal mirrorpolygonal mirror

Light emitted from the IP is Light emitted from the IP is collected by a fiber-optic bundle collected by a fiber-optic bundle and funneled into a photomultiplier and funneled into a photomultiplier tube (PMT)tube (PMT)

PMT converts VL into ePMT converts VL into e-- current current



Electronic signal output from PMT Electronic signal output from PMT input to an ADCinput to an ADC

Digital output from ADC storedDigital output from ADC stored Raster swept out by rotating Raster swept out by rotating

polygonal mirror and stage polygonal mirror and stage stepping motors produces I(t) into stepping motors produces I(t) into PMT which eventually translates PMT which eventually translates into the stored DV(x,y): into the stored DV(x,y): PMTPMT→→ADCADC→→RAMRAM

IP exposed to bright light to erase IP exposed to bright light to erase any remaining trapped eany remaining trapped e-- ( (~~50%)50%)

IP mechanically reinserted into IP mechanically reinserted into cassette ready for usecassette ready for use

200200m and 100m and 100m pixel size - m pixel size - (14”x17”: 1780x2160 and (14”x17”: 1780x2160 and 3560x4320, respectively)3560x4320, respectively)


IndirectIndirect Flat Panel Detectors Flat Panel Detectors

Use an intensifying screen Use an intensifying screen (CsI) to generate VL photons (CsI) to generate VL photons from an x-ray exposurefrom an x-ray exposure

Light photons absorbed by Light photons absorbed by individual array photodetectorsindividual array photodetectors

Each element of the array Each element of the array (pixel) consists of transistor (pixel) consists of transistor (readout) electronics and a (readout) electronics and a photodetector areaphotodetector area

The manufacture of these The manufacture of these arrays is similar to that used in arrays is similar to that used in laptop screens: thin-film laptop screens: thin-film transistors (TFT)transistors (TFT)


Charged-Coupled Devices (CCD)Charged-Coupled Devices (CCD)

Form images from visible lightForm images from visible light Videocams & digital camerasVideocams & digital cameras Each picture element (pixel) a Each picture element (pixel) a

photosensitive ‘bucket’photosensitive ‘bucket’ After exposure, the elements After exposure, the elements

electronically readout via ‘shift-electronically readout via ‘shift-and-read’ logic and digitizedand-read’ logic and digitized

Light focused using lenses or Light focused using lenses or fiber-opticsfiber-optics Fluoroscopy (II)Fluoroscopy (II) Digital cineradiography (II)Digital cineradiography (II) Digital biopsy system Digital biopsy system

(phosphor screen)(phosphor screen) 1K and 2K CCDs used1K and 2K CCDs used

c.f. Bushberg, et al. The Essential Physics of Medical c.f. Bushberg, et al. The Essential Physics of Medical Imaging, 2Imaging, 2ndnd ed., pp. 298-299. ed., pp. 298-299.

DirectDirect Flat Panel Detectors Flat Panel Detectors

Use a layer of photoconductive Use a layer of photoconductive material (e.g., material (e.g., αα-Se-Se) atop a TFT ) atop a TFT arrayarray

ee-- released in the detector layer released in the detector layer from x-ray interactions used to from x-ray interactions used to form the image directlyform the image directly

X-rayX-ray→→ee--→TFT → →TFT → ADCADC→→RAMRAM High degree of eHigh degree of e-- directionality directionality

through application of E fieldthrough application of E field Photoconductive material can be Photoconductive material can be

made thick w/o degradation of made thick w/o degradation of spatial resolutionspatial resolution

Photoconductive materialsPhotoconductive materials Selenium (Z=34)Selenium (Z=34) CdTe, HgICdTe, HgI22 and PbI and PbI22


Indirect Flat Panel Detector (for comparison)Indirect Flat Panel Detector (for comparison)

Direct Flat Panel DetectorDirect Flat Panel Detector

Thin-Film Transistors (TFT)Thin-Film Transistors (TFT)

After the exposure is complete After the exposure is complete and the eand the e-- have been stored in have been stored in the photodetection area the photodetection area (capacitor), rows in the TFT (capacitor), rows in the TFT are scanned, activating the are scanned, activating the transistor gatestransistor gates

Transistor source (connected Transistor source (connected to photodetector capacitors is to photodetector capacitors is shunted through the drain to shunted through the drain to associated charge amplifiersassociated charge amplifiers

Amplified signal from each Amplified signal from each pixel then digitized and storedpixel then digitized and stored

X-rayX-ray→→VLVL→→ee--→→ADCADC→→RAMRAMc.f. Bushberg, et al. The Essential Physics of Medical c.f. Bushberg, et al. The Essential Physics of Medical

Imaging, 2Imaging, 2ndnd ed., p. 301. ed., p. 301.

Resolution and Fill FactorResolution and Fill Factor

Dimension of detector element largely determines spatial resolutionDimension of detector element largely determines spatial resolution 200200m and 100m and 100m pixel size typicalm pixel size typical For dimension of ‘a’ mm - Nyquist frequency: FFor dimension of ‘a’ mm - Nyquist frequency: FN N = 1/2a= 1/2a If a = 100If a = 100m m → → FFN N = 5 cycle/mm= 5 cycle/mm Fill factor = (light sensitive area)/(detector element area)Fill factor = (light sensitive area)/(detector element area) Trade-off between spatial resolution and contrast resolutionTrade-off between spatial resolution and contrast resolution


Image Digitization and ProcessingImage Digitization and Processing

After acquisition and correction of raw data, the image is After acquisition and correction of raw data, the image is ready for display processing.ready for display processing.

The image data consists of a matrix of numbers. Each The image data consists of a matrix of numbers. Each pixel is one matrix point. Each gray scale is a digital pixel is one matrix point. Each gray scale is a digital value. value. For example: a matrix can have 1024 x 1024 pixels and each For example: a matrix can have 1024 x 1024 pixels and each

pixel will have a value from 0 to 1024. Each value is related to pixel will have a value from 0 to 1024. Each value is related to the radiation exposure which created that pixel.the radiation exposure which created that pixel.

Digital Storage of ImagesDigital Storage of Images

Usually stored as a 2D array Usually stored as a 2D array (matrix) of data, I(x,y): I(1,1), (matrix) of data, I(x,y): I(1,1), I(2,1), … I(n,m-1), I(n,m)I(2,1), … I(n,m-1), I(n,m)

Each minute region of the Each minute region of the image is called a image is called a pixelpixel (picture (picture element) represented by one element) represented by one value (e.g., digital value, gray value (e.g., digital value, gray level or Hounsfield unit)level or Hounsfield unit)

Typical matrices:Typical matrices: CT: 512x512x12 bits/pixelCT: 512x512x12 bits/pixel CR: 1760x2140x10 bits/pixelCR: 1760x2140x10 bits/pixel DR: 2048x2560x16 bits/pixelDR: 2048x2560x16 bits/pixel

c.f. Huang, HK. Elements of Digital Radiology, p. 8.c.f. Huang, HK. Elements of Digital Radiology, p. 8.

Image ProcessingImage Processing

Image data is scaled to present image with appropriate Image data is scaled to present image with appropriate gray scale (O.D.) values regardless of the actual gray scale (O.D.) values regardless of the actual radiation used to produce the image.radiation used to produce the image.

Image data is frequency enhanced around structures of Image data is frequency enhanced around structures of importance.importance. Process involves mathematical filters.Process involves mathematical filters.

Image data is display processed to give desired contrast Image data is display processed to give desired contrast and density.and density. Process involves re-mapping along a chosen display (“H&D”) Process involves re-mapping along a chosen display (“H&D”)

curvecurve

Generic Display ProcessingGeneric Display Processing

Different manufacturers may use different versions of Different manufacturers may use different versions of generic image processing methods.generic image processing methods. E.g. Musica, PtoneE.g. Musica, Ptone All describe means of scaling and modifying image appearance.All describe means of scaling and modifying image appearance.

Different manufacturers use different exposure Different manufacturers use different exposure indicators.indicators. E.g. EI, S, IgME.g. EI, S, IgM All describe the relationship between the exposure to the All describe the relationship between the exposure to the

detector and the pixel value.detector and the pixel value.

Generic Elements of Display ProcessingGeneric Elements of Display Processing

Exposure Recognition.Exposure Recognition. Adjust for high/low average exposureAdjust for high/low average exposure

Signal Equalization:Signal Equalization: Adjust regions of low/high signal valueAdjust regions of low/high signal value

Grayscale RenditionGrayscale Rendition Convert signal values to display valuesConvert signal values to display values

Edge Enhancement:Edge Enhancement: Sharpen edgesSharpen edges

M. Flynn, RSNA 1999M. Flynn, RSNA 1999

Image Processing Image Processing

normalized histogram comparison

0

0.005

0.01

0.015

0.02

0.025

0 0.2 0.4 0.6 0.8 1

normalized pixel value

frac

tio

n o

f p

ixel

s in

RO

Iph1.2

ptx case 4


IP dynamic range = 10IP dynamic range = 1044, about , about 100x that of S-F (10100x that of S-F (1022))

Very wide latitude Very wide latitude →→ flat contrast flat contrast Image processing required:Image processing required:

Enhance contrastEnhance contrast Spatial-frequency filteringSpatial-frequency filtering

CR’s wide latitude and image CR’s wide latitude and image processing capabilities produce processing capabilities produce reasonable OD or DV for either reasonable OD or DV for either under or overexposed examsunder or overexposed exams

Helps in portable radiography: Helps in portable radiography: where the tight exposure limits of where the tight exposure limits of S-F are hard to achieveS-F are hard to achieve

Underexposed Underexposed →→ ↑↑ quantum quantum mottle and overexposed mottle and overexposed →→ unnecessary patient doseunnecessary patient dose


Unsharpmasked Spatial Frequency ProcessingUnsharpmasked Spatial Frequency Processing


Global ProcessingGlobal Processing

Most common global image Most common global image processing: window/levelprocessing: window/level

Global processing algorithmGlobal processing algorithm I’(x,y) = c ∙ [I(x,y) – a]: I’(x,y) = c ∙ [I(x,y) – a]:

essentially y = mx + bessentially y = mx + b Level (brightness) set by aLevel (brightness) set by a Window (contrast) set by cWindow (contrast) set by c I’ = [2I’ = [2NN/ww]∙[I-{wl-(ww/2)}], /ww]∙[I-{wl-(ww/2)}],

where ww = window width and where ww = window width and wl = window levelwl = window level

Need threshold limits when Need threshold limits when max/min [2max/min [2NN-1, 0] digital values -1, 0] digital values encounteredencountered

If I’(x,y) > TIf I’(x,y) > Tmaxmax→→I’(x,y) = TI’(x,y) = Tmaxmax

If I’(x,y) < TIf I’(x,y) < Tminmin→→I’(x,y) = TI’(x,y) = Tminmin

c.f. Bushberg, et al. The Essential Physics of Medical c.f. Bushberg, et al. The Essential Physics of Medical Imaging, 2Imaging, 2ndnd ed., pp. 92 and 311. ed., pp. 92 and 311.

Image Processing Based on ConvolutionImage Processing Based on Convolution

Convolution: Ch. 10 - Image Convolution: Ch. 10 - Image Quality and Ch. 13 - CTQuality and Ch. 13 - CT

Defined mathematically as Defined mathematically as passing a N-dimensional passing a N-dimensional convolution kernel over an N-convolution kernel over an N-dimensional numeric array (e.g., dimensional numeric array (e.g., 2D image or CT transmission 2D image or CT transmission profile)profile)

At each location (x, y, z, t, ...) in At each location (x, y, z, t, ...) in the number array multiply the the number array multiply the convolution kernel values by the convolution kernel values by the associated values in the numeric associated values in the numeric array and sumarray and sum

Place the sum into a new numeric Place the sum into a new numeric array at the same locationarray at the same location



Delta function kernelDelta function kernel

Blurring kernel (normalization) Blurring kernel (normalization) also known as low-pass filteralso known as low-pass filter

Edge sharpening kernelEdge sharpening kernel


00 00 00

00 11 00

00 00 00

1/91/9 1/91/9 1/91/9

1/91/9 1/91/9 1/91/9

1/91/9 1/91/9 1/91/9

-1-1 -1-1 -1-1

-1-1 99 -1-1

-1-1 -1-1 -1-1


Convolution kernels can be much larger than 3 x 3, but Convolution kernels can be much larger than 3 x 3, but usually N x M with N and M oddusually N x M with N and M odd

Can also perform edge sharpening by subtracting Can also perform edge sharpening by subtracting blurred image from original blurred image from original → high-frequency detail → high-frequency detail (harmonization)(harmonization)

The edge sharpened image can then be added back to The edge sharpened image can then be added back to the original image to make up for some blurring in the the original image to make up for some blurring in the original image: CR unsharpmasking - freq. processingoriginal image: CR unsharpmasking - freq. processing

The effects of convolution cannot in general be undone The effects of convolution cannot in general be undone by a ‘de-convolution’ process due to the presence of by a ‘de-convolution’ process due to the presence of noise, but a deconvolution kernel can be applied to noise, but a deconvolution kernel can be applied to produce an approximation: produce an approximation: 1919F MRIF MRI

Median and Sigma FilteringMedian and Sigma Filtering

Convolution of an image with a kernel where all the Convolution of an image with a kernel where all the values are the same, e.g. (1/NxM), essentially performs values are the same, e.g. (1/NxM), essentially performs an average over the kernel footprintan average over the kernel footprint

Smoothing or noise reductionSmoothing or noise reduction This can make the resulting output value susceptible to This can make the resulting output value susceptible to

outliers (high or low)outliers (high or low) Median filter: rank order values in kernel footprint and Median filter: rank order values in kernel footprint and

take the median (middle) valuetake the median (middle) value Sigma filter: set sigma (Sigma filter: set sigma () value (e.g., 1) and throw out ) value (e.g., 1) and throw out

all values in kernel footprint > all values in kernel footprint > + + or < or < – – and then and then take the average and place in output imagetake the average and place in output image

Multiresolution/Multiscale Processing and Multiresolution/Multiscale Processing and Adaptive Histogram Equalization (AHE) Adaptive Histogram Equalization (AHE)

Some CR systems (Agfa/Fuji) make use of Some CR systems (Agfa/Fuji) make use of multiresolution image processing (AKA unsharpmasking) multiresolution image processing (AKA unsharpmasking) to enhance spatial resolutionto enhance spatial resolution

Wavelet or pyramidal processing on multiple frequency Wavelet or pyramidal processing on multiple frequency scalesscales

Histogram equalization re-distributes image digital values Histogram equalization re-distributes image digital values to uniformly span the entire digital value range [2to uniformly span the entire digital value range [2NN-1,0] to -1,0] to maximize contrastmaximize contrast

AHE does this on a spatial sub-region basis in an image AHE does this on a spatial sub-region basis in an image rather than the entire imagerather than the entire image

Fuji ‘Dynamic Range Control’ (DRC) a version of AHE Fuji ‘Dynamic Range Control’ (DRC) a version of AHE that operates on sub-regions of digital valuesthat operates on sub-regions of digital values

Histogram EqualizationHistogram EqualizationProperly Exposed ImageProperly Exposed Image Over-exposed ImageOver-exposed Image

Under-exposed ImageUnder-exposed Image Histogram Equalized ImageHistogram Equalized Image

c.f. http://www.wavemetrics.com/products/igorpro/imageprocessing/imagetransforms/histmodification.htmc.f. http://www.wavemetrics.com/products/igorpro/imageprocessing/imagetransforms/histmodification.htm

Global and Adaptive Histogram EqualizationGlobal and Adaptive Histogram Equalization

The following images illustrate The following images illustrate the differences between global the differences between global and adaptive histogram and adaptive histogram equalization.equalization.

MR image with the corresponding gray-scale MR image with the corresponding gray-scale histogram. The histogram has a peak at histogram. The histogram has a peak at minimum intensity consistent with the minimum intensity consistent with the relatively dark nature of the image. relatively dark nature of the image.

Global histogram equalization and the final Global histogram equalization and the final gray-scale histogram. Comparing the results gray-scale histogram. Comparing the results with the figure above we can see that the with the figure above we can see that the distribution was shifted towards higher values distribution was shifted towards higher values while the peak at minimum intensity remains.while the peak at minimum intensity remains.

Adaptive histogram equalization shows better contrast over different parts of the image. The corresponding gray-scale histogram lacks the mid-levels present in the global histogram equalization as a result of setting a high contrast level.

c.f. http://www.wavemetrics.com/products/igorpro/imageprocessing/imagetransforms/histmodification.htmc.f. http://www.wavemetrics.com/products/igorpro/imageprocessing/imagetransforms/histmodification.htm

Contrast vs. Spatial Resolution in Digital ImagingContrast vs. Spatial Resolution in Digital Imaging

S-F mammography can S-F mammography can produce images w/ > 20 lp/mmproduce images w/ > 20 lp/mm

According to Nyquist criterion According to Nyquist criterion would require 25 would require 25 m/pixel m/pixel resulting in a 7,200 x 9,600 resulting in a 7,200 x 9,600 image (132 Mbytes/image)image (132 Mbytes/image)

Digital systems have inferior Digital systems have inferior spatial resolutionspatial resolution

However, due to wide dynamic However, due to wide dynamic range of digital detectors and range of digital detectors and image processing capabilities, image processing capabilities, digital systems have superior digital systems have superior contrast resolutioncontrast resolution


Digital Imaging Systems and DQEDigital Imaging Systems and DQE

Remember the equation for Remember the equation for DQE(f)?DQE(f)?

DQE(f) = DQE(f) =

How can we account for this?How can we account for this? Both CR and the screens in Both CR and the screens in

film/screens made thinfilm/screens made thin Film higher spatial resolution Film higher spatial resolution

than CRthan CR DQE higher for DQE higher for αα-Si systems -Si systems

using CsI and Gdusing CsI and Gd22OO22S rather S rather than than αα-Se (mean x-ray E & Z)-Se (mean x-ray E & Z)

αα-Si DQE falling off more -Si DQE falling off more rapidly than rapidly than αα-Se (geometry)-Se (geometry)

2( )

( )

k MTF f

N NPS f

α-Si DR

α-Se DR

Digital versus Analog Processes & ImplementationDigital versus Analog Processes & Implementation

Although some of the previous image reception systems Although some of the previous image reception systems were labeled ‘digital’, the initial stage of those devices were labeled ‘digital’, the initial stage of those devices produce an analog signal that is later digitizedproduce an analog signal that is later digitized

CR: x-raysCR: x-rays→VL→PMT→current→voltage→ADC→VL→PMT→current→voltage→ADC CCD, direct & indirect digital detectors: stored eCCD, direct & indirect digital detectors: stored e- - → ADC→ ADC Benefits of CRBenefits of CR

Same exam process and equipment as screen-film radiography Same exam process and equipment as screen-film radiography Many exam rooms serviced by one readerMany exam rooms serviced by one reader Lower initial costLower initial cost

Benefits of DRBenefits of DR Throughput Throughput ↑↑: radiographs available immediately for QC & read: radiographs available immediately for QC & read

Patient Dose ConsiderationsPatient Dose Considerations

Over and underexposed digital receptors produce Over and underexposed digital receptors produce images with reasonable OD or gray scale valuesimages with reasonable OD or gray scale values

As overexposure can occur, need monitoring programAs overexposure can occur, need monitoring program CR IP acts like a 200 speed S-F system wrt. QDE CR IP acts like a 200 speed S-F system wrt. QDE Use the CR sensitivity (‘S’) number to track doseUse the CR sensitivity (‘S’) number to track dose

Bone, spine and extremities: 200Bone, spine and extremities: 200 Chest: 300Chest: 300 General imaging including abdomen and pelvis: 300/400General imaging including abdomen and pelvis: 300/400

Flat panel detectors can reduce radiation dose by 2-3x Flat panel detectors can reduce radiation dose by 2-3x as compared with CR for the same image quality due to as compared with CR for the same image quality due to ↑ quantum absorption efficiency & conversion efficiency↑ quantum absorption efficiency & conversion efficiency

S-number DashboardMain Exams

0

50

100

150

200

250

300

350

400

450

500

MFEM MC2 MC5 MCH2 MKUB MPELV

Exam Code

S-n

um

be

r

Sep-03 Baseline

Mar-04

Apr-04

Target Values

Fixed Chest - [255-345]

Bone, Spine & Ext -[170-230]

General Imaging incl. Abdomen - [340-460]

Using the CR Sensitivity Number to Track DoseUsing the CR Sensitivity Number to Track Dose

Huda Ch6: Digital X-ray Imaging Question Huda Ch6: Digital X-ray Imaging Question

1212. Photostimulable phosphor systems do NOT include:. Photostimulable phosphor systems do NOT include:

A. Analog-to-digital converters A. Analog-to-digital converters B. Barium fluorohalideB. Barium fluorohalide C. Light detectors (blue)C. Light detectors (blue) D. Red light lasersD. Red light lasers E. Video camerasE. Video cameras


1111. Which of the following x-ray detector materials emits . Which of the following x-ray detector materials emits visible light:visible light:

A. XenonA. Xenon B. Mercuric iodideB. Mercuric iodide C. Lead iodideC. Lead iodide D. SeleniumD. Selenium E. Cesium iodideE. Cesium iodide

Raphex 2002 Question: Digital RadiographyRaphex 2002 Question: Digital Radiography

D47. Concerning computed radiography (CR), which of the following is true?

A. Numerous, small solid-state detectors are used to capture the x-ray exposure patterns.

B. It has better spatial resolution than film. C. It is ideal for portable x-ray examinations, when

phototiming cannot be used. D. It is associated with high reject/repeat rates. E. The image capture, storage, and display are

performed by the receiver.


1313. Photoconductors convert x-ray energy directly into:. Photoconductors convert x-ray energy directly into:

A. LightA. Light B. CurrentB. Current C. HeatC. Heat D. ChargeD. Charge E. RF energyE. RF energy


1515. Which of the following does NOT involve image . Which of the following does NOT involve image processing:processing:

A. Background subtractionA. Background subtraction B. Energy subtractionB. Energy subtraction C. Histogram equalizationC. Histogram equalization D. K-edge filteringD. K-edge filtering E. Low-pass filteringE. Low-pass filtering


1414. Processing a digital x-ray image by unsharpmask . Processing a digital x-ray image by unsharpmask enhancement would increase the:enhancement would increase the:

A. Bit depth per pixelA. Bit depth per pixel B. Matrix sizeB. Matrix size C. Patient doseC. Patient dose D. Visibility of edgesD. Visibility of edges E. Limiting spatial resolutionE. Limiting spatial resolution

Adjuncts and other interesting stuffAdjuncts and other interesting stuff

Geometric (Linear) TomographyGeometric (Linear) Tomography

With the advent of CT, geometric With the advent of CT, geometric tomography has only limited tomography has only limited clinical utility where only one or a clinical utility where only one or a few planes of objects with high few planes of objects with high contrast are desired, e.g., IVPcontrast are desired, e.g., IVP

Desired slice through patient set at Desired slice through patient set at pivot point (focal plane)pivot point (focal plane)

The tomographic process blurs out The tomographic process blurs out regions outside the focal plane, regions outside the focal plane, but still contributes to overall loss but still contributes to overall loss of contrastof contrast

Larger tomographic angles result Larger tomographic angles result in a lessening of out of plane in a lessening of out of plane contributionscontributions

High dose, comparable to CT for High dose, comparable to CT for many tomographic slicesmany tomographic slices


Digital TomosynthesisDigital Tomosynthesis

Improved version of geometric Improved version of geometric tomography where a digital tomography where a digital detector saves an image at detector saves an image at each of several tube angleseach of several tube angles

This allows reconstruction of This allows reconstruction of multiple planes through the multiple planes through the object through shifting the object through shifting the various images through a various images through a certain distance before certain distance before summing themsumming them

Much more dose efficient, but Much more dose efficient, but still suffers from out of plane still suffers from out of plane blurring effectsblurring effects

Either CR or DR usedEither CR or DR used


Temporal SubtractionTemporal Subtraction

Digital Subtraction Angiography Digital Subtraction Angiography (DSA) – usually 1K resolution(DSA) – usually 1K resolution

Mask (background) subtracted Mask (background) subtracted from images during/post contrast from images during/post contrast injection: injection: ΔΔ < 1% trans. visualized < 1% trans. visualized

Motion can cause misregistration Motion can cause misregistration artifactsartifacts

Digital value proportional to Digital value proportional to contrast concentration and vessel contrast concentration and vessel thicknessthickness

IIss = ln(I = ln(Imm) – ln(I) – ln(Icc) = ) = vesselvessel ∙ t ∙ tvesselvessel

Temporal subtraction works best Temporal subtraction works best when time differences between when time differences between images is shortimages is short

Possible to spatially warp images Possible to spatially warp images taken over a longer period of time taken over a longer period of time


Dual-Energy SubtractionDual-Energy Subtraction

Exploits differences between Exploits differences between the Z of bone (Zthe Z of bone (Zeffeff ≈ ≈ 13) and 13) and

soft tissue (Zsoft tissue (Zeffeff ≈ ≈ 7.6)7.6) Images taken either at two Images taken either at two

different kVp (two-shot)different kVp (two-shot) One image (one-shot) taken One image (one-shot) taken

with energy separation with energy separation provided by a filter (sandwich)provided by a filter (sandwich)

IIoutout = log = logee(I(Ilowlow) – R ∙ log) – R ∙ logee(I(Ihighhigh), ),

where R is altered to produce where R is altered to produce soft-tissue predominant or soft-tissue predominant or bone predominant imagesbone predominant images

GE Chest DR @ SCCAGE Chest DR @ SCCAc.f. Bushberg, et al. The Essential Physics of Medical c.f. Bushberg, et al. The Essential Physics of Medical

Imaging, 2Imaging, 2ndnd ed., p. 324. ed., p. 324.

Dual-Energy SubtractionDual-Energy Subtraction



2222. The matrix size in a DSA image is typically:. The matrix size in a DSA image is typically:

A. 128 x 128A. 128 x 128 B. 256 x 256B. 256 x 256 C. 512 x 512C. 512 x 512 D. 1024 x 1024D. 1024 x 1024 E. 2048 x 2048E. 2048 x 2048


2525. Changing the DSA matrix from 1024. Changing the DSA matrix from 102422 to 2048 to 204822 would would NOT increase the:NOT increase the:

A. Data digitization rateA. Data digitization rate B. Data storage requirementB. Data storage requirement C. Image processing timeC. Image processing time D. Spatial resolutionD. Spatial resolution E. Pixel sizeE. Pixel size

Raphex 2003 Question: Digital RadiographyRaphex 2003 Question: Digital Radiography

D51. A flat panel digital radiographic detector has a square 20 x 20 cm image receptor field. The full field of the detector is coupled to a nominal 2048 x 2048 CCD array. The relative spatial resolution (lp/mm) when going from a 20 x 20 cm to a 10 x 10 cm field of view is:

A. Four times better B. Twice as good C. The same D. Half as good E. One fourth as good


1717. The Nyquist frequency for a 1K digital photospot . The Nyquist frequency for a 1K digital photospot image (25 cm image intensifier diameter) is:image (25 cm image intensifier diameter) is:

A. 1 lp/mmA. 1 lp/mm B. 2 lp/mm B. 2 lp/mm C. 4 lp/mmC. 4 lp/mm D. 8 lp/mmD. 8 lp/mm E. 10 lp/mmE. 10 lp/mm

FFNN (lp/mm) = 1/2a = 1/2(1024 lines/250 mm) = 2.048 ≈ 2 (lp/mm) = 1/2a = 1/2(1024 lines/250 mm) = 2.048 ≈ 2

Digital Representation of Data (1)Digital Representation of Data (1)

Bits, Bytes and WordsBits, Bytes and Words Smallest unit of storage capacity = 1 bit (Smallest unit of storage capacity = 1 bit (bbinary diginary digitit: 1 or 0): 1 or 0) Bits grouped into bytes: 8 bits = byteBits grouped into bytes: 8 bits = byte Word = 16, 32 or 64 bits, depending on the computer system Word = 16, 32 or 64 bits, depending on the computer system

addressing architectureaddressing architecture

Computer storage capacity is measured in:Computer storage capacity is measured in: kilobytes (kB) - 2kilobytes (kB) - 21010 bytes = 1024 bytes bytes = 1024 bytes a thousand bytes a thousand bytes megabytes (MB) - 2megabytes (MB) - 22020 bytes = 1024 kilobytes bytes = 1024 kilobytes a million bytes a million bytes gigabytes (GB) - 2gigabytes (GB) - 23030 bytes = 1024 megabytes bytes = 1024 megabytes a billion bytes a billion bytes terabytes (TB) - 2terabytes (TB) - 24040 bytes = 1024 gigabytes bytes = 1024 gigabytes a trillion bytes a trillion bytes

Digital Representation of Data (2)Digital Representation of Data (2)

Digital Representation of Different Types of DataDigital Representation of Different Types of Data Alphanumeric text, integers, and non-integer dataAlphanumeric text, integers, and non-integer data

Storage of Positive IntegersStorage of Positive Integers In general, n bits have 2In general, n bits have 2nn possible permutations and can possible permutations and can

represent integers from 0 to 2represent integers from 0 to 2nn-1 (the range usually denoted with -1 (the range usually denoted with square brackets):square brackets):

n bits represents 2n bits represents 2nn values with range [0, 2 values with range [0, 2nn-1]-1] 8 bits represents 28 bits represents 288 = 256 values with range [0, 255] = 256 values with range [0, 255] 10 bits represents 210 bits represents 21010 = 1024 values with range [0, 1023] = 1024 values with range [0, 1023] 12 bits represents 212 bits represents 21212 = 4096 values with range [0, 4095] = 4096 values with range [0, 4095] 16 bits represents 216 bits represents 21616 = 65,536 values with range [0, 65535] = 65,536 values with range [0, 65535]

Conversion of Analog Data to Digital FormConversion of Analog Data to Digital Form

The electronic measuring devices of medical scanners (e.g., The electronic measuring devices of medical scanners (e.g., transducers and detectors) produce analog signalstransducers and detectors) produce analog signals

Analog to digital conversion (analog to digital converter – ADC)Analog to digital conversion (analog to digital converter – ADC) ADCs characterized byADCs characterized by

sampling rate or frequency (e.g., samples/sec – 1 MHz)sampling rate or frequency (e.g., samples/sec – 1 MHz) number of bits output per sample (e.g., 12 bits/sample = 12-bit ADC)number of bits output per sample (e.g., 12 bits/sample = 12-bit ADC)

c. f. Bushberg, et al., The Essential Physics of Medical Imaging, 2nd ed., p. 69.c. f. Bushberg, et al., The Essential Physics of Medical Imaging, 2nd ed., p. 69.

Periodic Table of the ElementsPeriodic Table of the Elements

c.f. http://www.ktf-split.hr/periodni/en/c.f. http://www.ktf-split.hr/periodni/en/

digital radiography – chapter 11 adjuncts to radiology – chapter 12 brent k. stewart, phd, dabmp...

Documents

image processing

image receptor

digital image correction

final digital image

dr digital radiography

digital imaging

image quality

image data