368 - A Novel Material Decomposition Method for Dual-Energy and Single-Energy Computed Tomography
Presenter(s)
B. Huang1, H. C. Kuo1, M. F. Chan2, H. Xie2, W. Cai2, Y. Fu2, U. Mahmood1, N. Y. Lee3, P. Iyengar3, J. J. Cuaron3, L. I. Cervino2, J. M. Moran2, T. Li2, and X. Li2; 1Memorial Sloan Kettering Cancer Center, New York, NY, 2Department of Medical Physics, Memorial Sloan Kettering Cancer Center, New York, NY, 3Department of Radiation Oncology, Memorial Sloan Kettering Cancer Center, New York, NY
Purpose/Objective(s): Material decomposition is the backbone and foundation of Dual-Energy computed tomography (DECT) for disease diagnosis and radiotherapy applications. However, robust material decomposition remains challenging due to the inherited low soft tissue contrast and high noise level, especially in the low-kilo Voltage (KV) CT images. This study aims to introduce a novel material decomposition method that can be applied to both DECT and single-energy CT (SECT) modalities.
Materials/Methods: We applied a well-established mixture image segmentation framework to perform the material decomposition for DECT and SECT images. Specifically, the CT image intensity of a voxel is mathematically described by a mixture Gaussian model, where each tissue type is represented by a Gaussian distribution of intensity values. The overall probability of the observed intensity is the weighted sum of these distributions, with weights reflecting the tissue types and their proportions in the voxel. Furthermore, a Markov random field (MRF) model is employed to incorporate spatial context into the tissue classification: the tissue type assigned to each voxel is influenced by the tissue types of its neighboring voxels, ensuring spatial smoothness in the classification. Following the maximum-posterior probability (MAP) framework, we used the iterated conditional modes (ICM) optimization approach to estimate the material decomposition distributions. The proposed framework was tested on five patients who underwent CT scans with 100/140 KVp with tin filter settings and iodine-based contrast enhancement. For the DECT application, we calculated the material distribution maps of five different tissue types: Lung, fat, muscle, iodine, and calcium; for the 100 kVp SECT application, decomposition of four materials including lung, fat, muscle, and iodine/calcium was performed.
Results: We selected a total of 30 different regions of interest (ROI) from the five test patients, including liver tumors, blood vessels, and lymph nodes, and we calculated the contrast-to-noise ratio (CNR) of these ROIs on the original CT images as well as the iodine map images. Compared to the original CT images, the CNRs for the blood vessels and lymph nodes in the iodine map images were increased by 920.1%± 380.4%; for the liver tumors, the CNRs were increased by 71.5%±37.4%. For the fat, muscle, and iodine map images (within a known iodine region), the CNR differences were within 5% between the DECT and SECT-based material decomposition.
Conclusion: The proposed framework successfully decomposed the SECT or DECT images into four or five different material map images with significantly improved CNR. Material Decomposition used in the SECT image domain can extend some DECT functions to a much larger patient population without DECT, even though the iodine and calcium cannot be adequately separated.