Code related to the TDBVM network

This method adopts a dual encoder architecture to extract topological structure priors and deep semantic features separately. The topology branch introduces a multi-color space topology component extraction module to generate a topology information graph.

Through multi-scale fusion mechanism, topological priors effectively guide visual feature learning, enhance the model's perception ability of complex boundaries and lesion morphology, and effectively suppress artifact interference. The code supports model training, inference testing, and experimental reproduction, and can be used to verify the relevant experimental settings and results in the paper.