Room: Room 202
Purpose: Triple-energy CT (TECT) enhances tissue characterization because of its ability in material decomposition. However, the quality of material decomposition image will be degraded by a direct matrix inversion material decomposition. In this study, we present a strategy of noise suppression in image-domain material decomposition for TECT via the penalized weighted least-squares criteria with total variation regularization (PWLS-TV).
Methods: We formulated the proposed PWLS-TV method in the form of weighted least-squares with total variation regularization (TV). We first analyzed the statistical properties of the material decomposition, where the inverse of the estimated noise variance-covariance matrix of the material decomposition image was considered as the weight in the weighted least-squares term. The TV regularization enforces the material decomposition image smoothness by penalizing the differences of local neighbor pixels. We implemented the PWLS-TV method using an alternation minimization algorithm and evaluated the PWLS-TV method using phantom studies and compare the method with direct matrix inversion material decomposition.
Results: On a mouse thorax phantom, the PWLS-TV method outperforms the direct matrix inversion method in terms of noise suppression and structural information preservation. Compared to the direct matrix inversion method, the relative root mean square error (RRMSE) reduces by 34%, 26% and 21% in soft tissue, bone, and iodine image, respectively. The structural similarity (SSIM) value increases by 27%, 10% and 11% in soft tissue, bone and iodine image, respectively.
Conclusion: The PWLS-TV method can efficiently suppress the noise in material decomposition for TECT. The PWLS-TV method improves the material decomposition image quality of TECT as compared to direct matrix inversion material decomposition.
Funding Support, Disclosures, and Conflict of Interest: This Work is supported in party by the Cancer Prevention and Research Institute of Texas (RP160661), US National Institutes of Health (R01 EB020366), National Natural Science Foundation of China (11701097, U1708267), Natural Science Foundation of Jiangxi Province (2016BAB212055).