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

Area of Science:

• Biomedical Imaging
• Medical Diagnostics
• Computational Biology

Background:

• Dynamic fluorescence molecular tomography (DFMT) enables 3D monitoring of fast biological processes.
• Early detection of liver disease necessitates distinguishing normal from injured liver tissues.
• Current DFMT methods face challenges due to ill-posed problems and signal interference, limiting liver injury monitoring.

Purpose of the Study:

• To develop a novel method for distinguishing normal and injured liver tissues using DFMT.
• To enhance early liver disease monitoring and localization capabilities.
• To overcome signal interference issues in DFMT for liver applications.

Main Methods:

• A Signal Separation Net (SSN) was proposed, integrating Convolutional Long Short-Term Memory (ConvLSTM) and ResNet architectures.
• ConvLSTM was utilized to separate liver injury projections from DFMT surface photon distribution.
• ResNet established the nonlinear relationship between projections and liver injury localization.

Main Results:

• The proposed SSN demonstrated effective separation of liver injury signals.
• The method achieved promising performance in liver injury monitoring and localization.
• Experimental findings validated the SSN's viability for DFMT-based liver disease assessment.

Conclusions:

• The novel SSN is a viable tool for early liver injury monitoring using DFMT.
• This approach shows significant potential for advancing non-invasive liver disease diagnostics.
• The SSN effectively addresses signal separation challenges in DFMT for improved clinical translation.