Encrypted login | home

Program Information

Sparse Dictionary Method for Material Elemental Decomposition in Dual and Multi-Energy CT for Proton Stopping Power Ratio Estimation


C Shen

C Shen1*, B Li1 , L Chen1 , J Wang1 , M Yang1 , Y Lou2 , X Jia1 , (1) University of Texas Southwestern Medical Center, Dallas, TX, (2) University of Texas as Dallas, Dallas, Texas

Presentations

TH-AB-605-10 (Thursday, August 3, 2017) 7:30 AM - 9:30 AM Room: 605


Purpose: Estimation of stopping power ration (SPR) relative to water in human tissue is crucial to proton therapy to reduce range uncertainty. SPR depends on both relative electron density to water (rED) and elemental composition (EC). Hence, large SPR uncertainty exists in current standard practice that derives SPR using a single energy CT image. Dual-energy (DECT) has been recently shown to improve SPR estimation accuracy. In this study, we propose a new and unified framework for both DECT and multi-energy CT (MECT) to determine SPR.

Methods: We first constructed a dictionary containing material compositions of typical tissues existing in human. We formulated the decomposition problem as an optimization problem that minimized the difference between measured and calculated CT numbers in different energy channels with respect to rED and EC. We also enforced that the solution EC is a sparse combination of the ECs of those materials in the dictionary. The optimization problem was non-convex and was solved by a two-loop iterative algorithm that alternatively optimize rED and EC. After solving the optimization problem, SPR of the material was calculated using Bethe-Bloch equation. We tested our method in both simulation and experimental studies. We also studied robustness of our method with respect to parameter selection.

Results: The root-mean-square (RMS) errors of SPR averaged over all the testing materials were 0.28%, 0.21% and 0.14% in simulation study for DECT, MECT (three energy channels) and MECT (four channels.) The average RMS errors in real experiments performed on a Gammax phantom were 0.70% 0.54% and 0.46% for the three settings, respectively. The algorithm was found to be robust to sparsity parameter and initial guess.

Conclusion: The proposed method can accurately calculate SPR using DECT and MECT. It is beneficial to use more energy channels for SPR calculation.


Contact Email: