SSN: Monitoring liver injure through signal separation network in dynamic fluorescence molecular tomography

Abstract

Dynamic fluorescence molecular tomography (DFMT) is a promising molecular imaging technique that offers the potential to monitor fast kinetic behaviors within small animals in three dimensions. Early monitoring of liver disease requires the ability to distinguish and analyze normal and injured liver tissues. However, the inherent ill-posed nature of the problem and energy signal interference between the normal and injured liver regions limit the practical application of liver injury monitoring. In this paper, a novel Signal Separation Net (SSN) is proposed to distinguish normal and injured liver tissue for early liver injury monitoring and liver injury localization. By employing a Convolutional Long Short-Term Memory (ConvLSTM) model, SSN first separates the projections of liver injury from the surface photon distribution of DFMT. Then, a ResNet is employed to establish the nonlinear relationship between the liver injury projections and liver injury localization. Experimental findings underscore the viability of the proposed SSN in achieving promising performance in liver injury monitoring using DFMT, demonstrating significant potential for advancing early liver disease monitoring.