MPU-Net: Multi-decoder U-Net Based on Prior Information for Multi-material Decomposition in Spectral CT

The material decomposition of spectral CT is the basis of clinical applications. In order to make full use of the prior information of basis material images, we proposed the MPU-Net to perform image-based material decomposition. The model has an encoder for feature sharing, and different decoders focus on the differences between the basis material images. To further improve the accuracy of this model, the basis material images’ correlation was extracted using HRNet and used to constrain the material decomposition process. The experimental results showed that the MPU-Net performed better than the reference methods in both visual evaluation results and quantitative indicators, demonstrating its effectiveness in improving the accuracy of material decomposition.