Encrypted login | home

Program Information

Prior Segmentation Assisted Statistical Multi-Material Decomposition for Dual-Energy CT

no image available
Y Xue

Y Jiang , C Yang , Q Lyu , K Sheng , Y Xue*, T Niu , UCLA School of Medicine, Los Angeles, CA


SU-F-201-6 (Sunday, July 30, 2017) 2:05 PM - 3:00 PM Room: 201

Purpose: Dual-energy CT (DECT) improves material differentiation because its capacity of material decomposition. Nevertheless, the decomposed image quality is severely degraded due to the biased pixel value and magnified noise in the decomposition process, which limits the quantitative DECT application. We propose a multi-material decomposition (MMD) method using prior segmentation as an assistance to tackle the above problems.

Methods: A direct decomposition method sequentially decomposes each pixel into different triplet and collects solution that satisfies a box ([0, 1]) and the sum-to-one constraints. It chooses the solution according to triplet priority list when multiple solutions exist. Nevertheless, an optimal list is hard to be acquired due to the short of information on the material property of current pixel. To solve the problem, we apply segmentation on the CT image before decomposition to coarsely locate the material distribution. The location information is used to select material triplet and thus pixel bias value is corrected. After the triplet selection, we apply a statistical MMD method to suppress the decomposition image noise similar to our previous work. The proposed method is evaluated on the CT image of a phantom at the energy of 75kVp and 125kVp.

Results: After the triplet selection, the decomposition accuracy is improved by 4.13% as compared with direct inversion method. After statistical MMD, the noise is significantly suppressed as compared with direct inversion method. The decomposition accuracy is improved by 14.92% and 4.13% compared to direct inversion method and statistical MMD method.

Conclusion: Effective bias correction of decomposed image is achieved using our segmentation assist decomposition method and the noise is significantly suppressed. The increased accuracy of decomposed image substantially facilitates DECT-based clinical applications, such as liver-fat quantification. As such, the proposed method will be very attractive in current DECT application.

Funding Support, Disclosures, and Conflict of Interest: Zhejiang Provincial Natural Science Foundation of China (Grant No. LR16F010001) National High-tech R&D Program for Young Scientists by the Ministry of Science and Technology of China (863 Program, 2015AA020917)

Contact Email: