A Lightweight Network for Contextual and Morphological Awareness for Hepatic Vein Segmentation

Area of Science:

• Medical Imaging
• Computer Vision
• Artificial Intelligence

Background:

• Accurate segmentation of the hepatic vein is crucial for liver disease diagnosis and treatment.
• The complex morphology and sparse distribution of hepatic veins make automatic segmentation challenging.
• Current methods struggle with precise segmentation due to data labeling difficulties.

Purpose of the Study:

• To develop an efficient and accurate automatic segmentation method for hepatic veins.
• To address the challenges posed by the small target size and diverse morphology of hepatic veins.
• To improve the precision of liver disease diagnosis and treatment through enhanced segmentation.

Main Methods:

• Proposed a lightweight contextual and morphological awareness network.
• Introduced a novel morphology aware module using an attention mechanism to enhance inter-slice continuity.
• Incorporated a 3D reconstruction module to leverage 3D contextual information with minimal parameters.

Main Results:

• The proposed method demonstrated enhanced segmentation accuracy (Dice coefficient) compared to state-of-the-art methods on two datasets.
• The network achieved superior performance with a significantly smaller number of parameters.
• The approach showed potential for stronger generalization capabilities.

Conclusions:

• The developed lightweight network offers an effective solution for automatic hepatic vein segmentation.
• The novel modules improve the handling of complex venous structures and 3D context.
• Reduced parameter count facilitates clinical deployment by lowering hardware requirements and enhancing generalizability.