United adversarial learning for liver tumor segmentation and detection of multi-modality non-contrast MRI

Simultaneous segmentation and detection of liver tumors (hemangioma hepatocellular carcinoma (HCC)) by using multi-modality non-contrast magnetic resonance imaging (NCMRI) are crucial for the clinical diagnosis. However, it is still a challenging task due to: (1) HCC information on NCMRI insufficient makes extraction feature difficult; (2) diverse characteristics in causes fusion selection (3) no specific between hemangioma cause difficult. In this study, we propose united adversarial learning framework (UAL) simultaneous NCMRI. The UAL first utilizes multi-view aware encoder to extract tumor detection. encoder, novel edge dissimilarity pyramid module designed facilitate complementary extraction. Secondly, newly channel used fuse make decision selection. Then, proposed mechanism coordinate sharing with padding integrates multi-task so that enables perform one discriminator. Lastly, an innovative multi-phase radiomics guided discriminator exploits clear improve performance via strategy. validated corresponding (i.e. T1FS pre-contrast MRI, T2FS DWI) three phases contrast-enhanced MRI 255 subjects. experiments show gains high dice similarity coefficient 83.63%, pixel accuracy 97.75%, intersection-over-union 81.30%, sensitivity 92.13%, specificity 93.75%, 92.94%, which demonstrate has great potential diagnosis tumors.