Transmission Tomographic Image Reconstruction using Optimization Transfer: Application to Joint Attenuation / Activity Image Reconstruction for TOF PET
Winter School in Imaging Science, A3 Inverse Problem Meeting
Abstract
Many state-of-the-art image reconstruction algorithms for low dose CT have used weighted least square based data fidelity terms. However, desiring much lower radiation dose for CT and improving energy-resolution in spectral CT detectors may lead to lower signal-to-noise ratio (SNR) of measurements (per energy bin). Low SNR data usually requires Poisson model based data fidelity terms for accurate image reconstruction. In this talk, we will review optimization transfer methods for transmission tomography with Poisson noise model, especially focusing on separable quadratic surrogate (SQS) algorithm. Then, we will show one example for using the SQS algorithm in our recent work on joint attenuation and activity image reconstruction from TOF PET data using alternating direction method of multipliers framework.
Publisher
A3 Foresight Program: Modeling and Computation of Applied Inverse Problems Medical Image Computing Group