Midterm Review. Yao Wang Polytechnic University, Brooklyn, NY 11201

Transcription

1 Midterm Review Yao Wang Polytechnic University, Brooklyn, NY Based on J. L. Prince and J. M. Links, Medical maging Signals and Systems, and lecture notes by Prince. Figures are from the textbook.

2 Topics to be Covered mage Quality (Chap3) Radiation Physics (Chap4) Projection Radiography (Chap5) Computed Tomography (Chap6) EL582 Radiation Physics Yao Wang, Polytechnic U., Brooklyn 2

3 mage Quality (Chap3, Lect.1) Contrast and modulation transfer function Resolution Noise: signal to noise ratio Diagnostic accuracy EL582 Radiation Physics Yao Wang, Polytechnic U., Brooklyn 3

4 Contrast Contrast: Difference between image characteristics of an object of interest and surrounding objects or background Global contrast Local contrast EL582 Radiation Physics Yao Wang, Polytechnic U., Brooklyn 4

5 EL582 Radiation Physics Yao Wang, Polytechnic U., Brooklyn 5 Modulation Transfer Function The actual signal being imaged can be decomposed into many sinusoidal signals with different frequencies Suppose the imaging system can be considered as a LS system with frequency response H(u,v) maged signal is The MTF refers to the ratio of the contrast (or modulation) of the imaged signal to the contrast of the original signal at different frequencies A H B v u H m y v x u B v u H A H y x g k k k k g k k k k k k (0,0) ), ( ); 2 sin(2 ), ( (0,0) ), (, + + π π A B m y v x u B A y x f k k f k k k k + +, ); 2 sin(2 ), ( π π (0,0) ), ( ), (,,,, H v u H m m v u MTF v u f v u g

6 MTF vs. Frequency Response MTF characterizes how the contrast (or modulation) of a signal component at a particular frequency changes after imaging MTF magnitude of the frequency response of the imaging system (normalized by H(0,0)) Typically 0 MTF( u, v) MTF(0,0) 1 Decreasing MTF at higher frequencies causes the blurring of high frequency features in an image EL582 Radiation Physics Yao Wang, Polytechnic U., Brooklyn 6

7 Resolution defined by FWHM Resolution refers to the ability of a system to depict spatial details. Resolution of a system can be characterized by its line spread function How wide a very thin line becomes after imaging Full width at half maximum (FWHM) determines the distance between two lines which can be separated after imaging The smaller is FWHM, the higher is the resolution Line spread function (impulse response) EL582 Radiation Physics Yao Wang, Polytechnic U., Brooklyn 7

8 Distance > FWHM Distance > FWHM Distance FWHM (barely separate) Distance < FWHM (cannot separate) EL582 Radiation Physics Yao Wang, Polytechnic U., Brooklyn 8

9 Resolution Defined by Bar Phantom The resolution of an imaging system can be evaluated by imaging a bar phantom. The resolution is the frequency (in lp/mm) of the finest line group that can be resolved after imaging. Gamma camera: 2-3 lp/cm CT: 2 lp/mm chest x-ray: 6-8 lp/mm EL582 Radiation Physics Yao Wang, Polytechnic U., Brooklyn 9

10 Resolution and MTF (or frequency response) A pure vertical sinusoidal pattern can be thought of as blurred image of uniformly spaced vertical lines The distance between lines is equal to distance between maxima f the frequency u 0, the distance 1/ u 0 f ( x, y) A + g( x, y) B sin(2πu H (0,0) A + H (0,0) A + 0 H ( u x) 0 MTF( u,0) sin(2πu,0) H (0,0) sin(2πu f the maximum frequency at which MTF(u) (or H(u)) is non-zero is u c, the minimum distance between distinguishable lines (i.e. FWHM) 1/ u c (u c is called the cut-off frequency of the system) The higher is the cut-off frequency, the higher is the system resolution 0 0 x) 0 x) EL582 Radiation Physics Yao Wang, Polytechnic U., Brooklyn 10

11 Example Which system below has better contrast and resolution? EL582 Radiation Physics Yao Wang, Polytechnic U., Brooklyn 11

12 Noise Noise refers to random fluctuation of signals in an image n x-ray imaging (projection or tomography), one source of noise is the random fluctuation in the number of photons generated and detected by the x-ray generator (N) N is modeled by Poisson distribution: Pr{ N ij k} k a k! e a ; k 0,1,... E{ N ij k} a Var{ N ij k} a Noise in the reconstructed image is a function of N, depending on the relation between the pixel value, the absorption properties of the imaged object, and N Differing in projection radiography and CT EL582 Radiation Physics Yao Wang, Polytechnic U., Brooklyn 12

13 Signal to Noise Ratio Amplitude SNR vs. Power SNR SNR in db SNR (db) 20 log 10 SNR a 10 log 10 SNR p EL582 Radiation Physics Yao Wang, Polytechnic U., Brooklyn 13

14 Contingency Table Diagnostic Accuracy EL582 Radiation Physics Yao Wang, Polytechnic U., Brooklyn 14

15 f the diagnosis is based on a single value of a test result and the decision is based on a chosen threshold, the sensitivity and specificity can be visualized as follows Should be able to determine a,b,c,d given the two distributions EL582 Radiation Physics Yao Wang, Polytechnic U., Brooklyn 15

16 Physics of Radiography (Chap4, Lect 2) onizing radiation and generation of x-ray Absorption and scattering of x-ray Radiation measurement EL582 Radiation Physics Yao Wang, Polytechnic U., Brooklyn 16

17 Radiation Physics onization: ejection of an orbiting electron from an atom, the affected atom produces radiation in the process of returning to ground state Two types of ionizing radiation Particulate (used to generate x-ray) EM (used for x-ray imaging) Particulate radiation transfers energy via Collisional transfer: resulting in heat Radioactive transfer: resulting in characteristic x-ray and Bremsstrahlung x-ray X-ray is produced by energetic electrons accelerated in a x-ray tube, consisting primarily Bremsstrahlung x-ray EM radiation transfers energy via Photoelectric effect: incident photons completely absorbed (good effect) Compton scattering: incident photons are deflected (bad effect) EL582 Radiation Physics Yao Wang, Polytechnic U., Brooklyn 17

18 Attenuation of EM Radiation Linear attenuation coefficient is the fraction of photons that are lost per unit length Depends on material property and the incident photon energy Bone, soft tissue, blood, etc. have very different attenuation property --- basis of x-ray imaging Fundamental photon attenuation law Homogeneous slab Heterogeneous slab Measured signal Reconstructed property EL582 Radiation Physics Yao Wang, Polytechnic U., Brooklyn 18

19 Measure of Radiation Effect Exposure: number of ions produced in a unit volume (Roentgen or R) Dose: absorbed energy per unit volume (rad) Exposure vs. dose: DfX n soft tissue, f~1: 1 Roenden exposure in soft tissue produces 1 rad exposure (f1) Equivalent dose: HDQ Q 1 for x-ray, gamma ray, electron, beta particle Q 10 for neutrons and protons Q 20 for alpha particles Effective dose: D w effective j organs w j H :weighting factor for organ j j EL582 Radiation Physics Yao Wang, Polytechnic U., Brooklyn 19

20 Projection Radiography (Chap5, Lec3) nstrumentation X-ray tube configuration Filtration and restriction of x-ray photons Compensation and Scatter control Film screen detector mage formation Basic imaging equation Geometric effect Extended source Detector/film response mage quality Contrast and SNR Effect of noise and Compton scattering EL582 Radiation Physics Yao Wang, Polytechnic U., Brooklyn 20

21 Radiographic System EL582 Radiation Physics Yao Wang, Polytechnic U., Brooklyn 21

22 X-ray Tube EL582 Radiation Physics Yao Wang, Polytechnic U., Brooklyn 22

23 X-Ray Spectra EL582 Radiation Physics Yao Wang, Polytechnic U., Brooklyn 23

24 Filtration, Restriction and Scatter Control and Contrast Agents Filtration: remove low energy photons (cannot differentiate among different tissues) Restriction and Scatter control: keeping only photons in straight line path Contrast agents: to change the attenuation coefficients of a body organ so that it can be imaged more clearly Dual energy X-ray imaging EL582 Radiation Physics Yao Wang, Polytechnic U., Brooklyn 24

25 Basic maging Equations EL582 Radiation Physics Yao Wang, Polytechnic U., Brooklyn 25

26 Example Blood vessel µ0.2 w/contrast µ20 Soft tissue µ0.4 w/o contrast : b o min max 0 0 (0.4*2.0) (0.4* *0.5) Global contrast : C e e Local contrast :C l ; o b b max max ; + min min 1) What is the local contrast of the blood vessel? 2) What is the local contrast of the blood vessel when contrast agent is injected? EL582 Radiation Physics Yao Wang, Polytechnic U., Brooklyn 26

27 maging of a Uniform Slab EL582 Radiation Physics Yao Wang, Polytechnic U., Brooklyn 27

28 EL582 Radiation Physics Yao Wang, Polytechnic U., Brooklyn 28

29 maging of a non-homogeneous thin slice and with extended source Md/z m1-m-(d-z)/z x: detector position x/m: source position x/m: object position EL582 Radiation Physics Yao Wang, Polytechnic U., Brooklyn 29

30 Overall maging Equation ncluding all effects (geometric, extended source, film-screen blurring), the image corresponding to a slab at z with transmittivity function t z (x,y) is Oblique and inverse square law source distribution t z (x,y)exp{-µ(x,y) z} Film blurring function The measured signal can be considered the input signal (t(x,y)) convolved with a combined impulse response depending on s(x,) and h(x,y) (source distribution and film response) EL582 Radiation Physics Yao Wang, Polytechnic U., Brooklyn 30

31 SNR in Projection Radiography C is the contrast in terms of absorption coefficient Assuming incoming photon follows the poisson distribution with mean sqrt(nb) EL582 Radiation Physics Yao Wang, Polytechnic U., Brooklyn 31

32 EL582 Radiation Physics Yao Wang, Polytechnic U., Brooklyn 32 Effect of Compton Scattering Compton scattering causes the incident photons to be deflected from their straight line path Add a constant intensity s in both target and background intensity ( fog ) Decrease in image contrast Decrease in SNR b b b b b t b t N C C C σ SNR contrast background intensity : target intensity : W/o scattering : b s b s b s b b b b b s s b b s b b t s b s t N N N C N N N C C C C C / 1 1 SNR / 1 SNR' 1 ' contrast background intensity : target intensity : W/ scattering : ' σ

33 Summary Projection radiography system consists of an x-ray tube, devices for beam filtration and restriction, compensation filters, grids, and a filmscreen detector (or digital detector, filmless) The detector reading (or image gray level) is proportional to the number of unabsorbed x-ray photons arriving at the detector, which depends on the overall attenuation in the path from the source to the detector The above relation must be modified to take into account of inverse square law, obliquity, anode heel effect, extended source and detector impulse response The degree of film darkening is nonlinearly related to the film exposure (detected x-ray) by the H&D curve Both detector noise and Compton scattering reduce contrast and SNR of the formed image EL582 Radiation Physics Yao Wang, Polytechnic U., Brooklyn 33

34 Computed Tomography (Chap6, Lec.4,5) Tomography vs. Projection Parallel vs. Fan Beam Projection Reconstruction from parallel projection Reconstruction from fanbeam projection Blur due to non-ideal detector and windowing Noise and SNR EL582 Radiation Physics Yao Wang, Polytechnic U., Brooklyn 34

35 Projection: Projection vs. Tomography A single image is created for a 3D body, which is a shadow of the body in a particular direction (integration through the body) EL582 Radiation Physics Yao Wang, Polytechnic U., Brooklyn 35

36 Projection vs. Tomography Tomography A series of images are generated, one from each slice of a 3D object in a particular direction (axial, coronal, sagittal) To form image of each slice, projections along different directions are first obtained, images are then reconstructed from projections (backprojection, Radon transform) EL582 Radiation Physics Yao Wang, Polytechnic U., Brooklyn 36

37 1 st Generation CT: Parallel Projections EL582 Radiation Physics Yao Wang, Polytechnic U., Brooklyn 37

38 3G: Fan Beam EL582 Radiation Physics Yao Wang, Polytechnic U., Brooklyn 38

39 CT Measurement Model EL582 Radiation Physics Yao Wang, Polytechnic U., Brooklyn 39

40 EL582 Radiation Physics Yao Wang, Polytechnic U., Brooklyn 40

41 CT Number Reconstructed image pixel value is CT number (12 bits) EL582 Radiation Physics Yao Wang, Polytechnic U., Brooklyn 41

42 Projection Slice Theorem Projection Slice theorem The Fourier Transform of a projection at angle θ is a line in the Fourier transform of the image at the same angle. EL582 Radiation Physics Yao Wang, Polytechnic U., Brooklyn 42

43 Reconstruction Algorithm for Parallel Projections Backprojection: Backprojection of each projection Sum Filtered backprojection: FT of each projection Filtering each projection in frequency domain nverse FT Backprojection Sum Convolution backprojection Convolve each projection with the ramp filter Backprojection Sum EL582 Radiation Physics Yao Wang, Polytechnic U., Brooklyn 43

44 Practical mplementation Projections g(l, θ) are only measured at finite intervals lnτ; τ chosen based on maximum frequency in G(ρ,θ), W 1/ >2W or <1/2W (Nyquist Sampling Theorem) W can be estimated by the number of cycles/cm in the projection direction in the most detailed area in the slice to be scanned For filtered backprojection: Fourier transform G(ρ,θ) is obtained via FFT using samples g(nτ, θ) f N sample are taken, 2N point FFT is taken by zero padding g(nτ, θ) For convolution backprojection The ramp-filter is sampled at lnτ Sampled Ram-Lak Filter c( n) 2 1/ 4τ ; 1/ ( nπτ ) 0; 2 ; n 0 n odd n even EL582 Radiation Physics Yao Wang, Polytechnic U., Brooklyn 44

45 Fan Beam: Equiangular Ray EL582 Radiation Physics Yao Wang, Polytechnic U., Brooklyn 45

46 Equiangular Ray Reconstruction γ D φ r β φ Reconstructed image is represented in the polar coordinate using ( r, φ). The relative position of a pixel at ( r, φ) to the source at ( D, β ) is specified by ( D', γ ') : D' 2 ( r, φ) tanγ '( r, φ) f ( r, φ) q( γ, β ) p'( γ, β ) 2 ( D + r sin( β φ) ) + ( r cos( β φ) ) ( r cos( β φ) ) ( D + r sin( β φ) ) Re construction formular : 2π 0 1 ( D' ) 2 p'( γ, β )* c q( γ ', β ) dβ ( γ ) p( γ, β ) D cos( γ ) f Weighted backprojection 2 1 γ c f ( γ ) c( γ ) 2 sin γ c( γ ) is the ramp filter used in parallel projection. 2 Derivation not required for this class. Detail can be found at [Kak&Slaney]. Note typos in [Prince&Links]. NOT required in exam. EL582 Radiation Physics Yao Wang, Polytechnic U., Brooklyn 46

47 Blur due to Non-ideal Detector and Ramp Filter Area detector: can be characterized by an indicator function s(l) (aks impulse response) Recall that the ideal filter c(ρ) is typically modified by a window function W(ρ) Overall effect on projection: g(l) g(l)*s(l)*w(l) Overall effect on reconstructed image: f (x,y)f(x,y)*h(x,y) h(x,y)r^-1{s(l)*w(l)} H^-1{ S(q)W(q)} Should know how to derive the blurring function based on given Hankel transform pair or from 2D Fourier transform directly EL582 Radiation Physics Yao Wang, Polytechnic U., Brooklyn 47

48 SNR of reconstructed image Noise is in detected number of photons for one source/detector pair, N_ij, which follows Poisson distribution, with mean variance\bar N_ij Projection data g_ij is a function of N_ij, with mean and variance each a function of \bar N_ij Reconstructed image is a function of g_ij for all i,j, filter cutoff (ρ 0 ), and detector width (T). With various assumptions and \bar N_ij\bar N, we get With ρ 0 1/T For fan-beam: N_f photon/fan, D detector/fan, L length of detector array EL582 Radiation Physics Yao Wang, Polytechnic U., Brooklyn 48

49 Midterm Exam 10/30/06, 3:35-5:50, RH503 Closed book, 1 sheet of notes allowed (double-sided) Covers chap3,4,5,6 Not covered: Matlab implementation Projection radiography: effects considering extended source and film impulse response will not be covered Fanbeam reconstruction algorithms Fourier transform pair and Hankel transform pair will be given if needed Office hours Thursday (10/26) 4-5, Monday? EL582 Radiation Physics Yao Wang, Polytechnic U., Brooklyn 49