Transcript Document
Tomographic Image Reconstruction Miljenko Markovic Overview • Image creation • Image reconstruction – Brute force – Iterative techniques – Backprojection – Filtered backprojection Image Creation • Tomogram – image of a slice taken through a 3D volume detector x-ray source collimator object • Projection – Attenuation profile through the object – The projection function represents the summation of the attenuation coefficients along a given X-ray path Image Creation • Sinogram – 2D data set – result of stacking all the projections together • Radon transform – Transformation of a function (image) into the sinogram, p(r) – Computes projections of an image along specified directions Image Reconstruction • Process of estimating an image from a set of projections • Several algorithms exist to accomplish this task: – – – – Brute force Iterative techniques Backprojection Filtered backprojection Brute Force • projection set defines a system of simultaneous linear equations - can be solved using algorithms from linear algebra • not practical for real systems (can have hundreds of simultaneous equations for a single slice) Iterative Reconstruction • Known as algebraic reconstruction technique – ART, consists of three steps: – Make an initial guess at the solution – Compute projections based on the guess – Refine the guess based on the weighted difference between the actual projections and the desired projections – Original reconstruction method used in medical imaging – Works, but is slow and susceptible to noise Backprojection • Propagates sinogram back into the image space along the projection paths (inverse Radon transform) • Backprojection image is a blurred version of the original image • The projection theorem (central slice theorem) - provides an answer to inverse Radon transform problem – Set of 1D Fourier transform of the Radon transform of a function is the 2D Fourier transform of that function Fourier Reconstruction • Calculate the 1D Fourier transform of all projections [p(r) = P(k)] • Place P(k) on polar grid to get P(k,) • Resample in Cartesian space to get F(kx,ky) • Calculate the 2D inverse Fourier transform of F(kx,ky) to get f(x,y) – image • Resultant image is noisy Fourier Reconstruction Filtered Backprojection • • • • • Take projections - sinogram Transform data to the frequency domain Filter data Inverse transform – smoothed sinogram Backproject Filtered Backprojection Filtered Backprojection 1. ramp filter + nearest neighbor algorithm 2. ramp & Hamming filter + nearest neighbor algorithm 3. ramp filter + linear interpolation 4. ramp & Hamming filter + linear interpolation 1 2 3 4 References • Image Processing: The Core of Nuclear Cardiology, Scott M. Leonard, MS, CNMT, Northwestern University, ppt presentation • Xiang Li , Jun Ni and Ge Wang, Parallel iterative cone beam CT image reconstruction on a PC cluster, Journal of X-Ray Science and Technology 13 (2005) 63–72 • HARISH P. HlRlYANNAlAH, X-ray Computed Tomography for Medical Imaging, IEEE SIGNAL PROCESSING MAGAZINE Thank you