HPA-UNet: A Hybrid Post-Processing Attention U-Net for Tongue Segmentation

Area of Science:

• Computer Vision
• Medical Imaging
• Traditional Chinese Medicine (TCM)

Background:

• Tongue diagnosis is a crucial Traditional Chinese Medicine (TCM) method for assessing health status.
• Accurate tongue segmentation is vital for computer-aided disease recognition but is challenging with existing methods.

Purpose of the Study:

• To develop a highly precise tongue segmentation method using computer vision.
• To improve the accuracy of automated disease detection based on tongue images.

Main Methods:

• A novel approach combining an improved U-shaped neural network with an edge refinement post-processing technique.
• Implementation of a data augmentation strategy to prevent network over-fitting.
• Development of a U-shaped neural network specifically designed for high-precision tongue image segmentation.

Main Results:

• The proposed method achieved competitive performance across two datasets.
• The edge refinement post-processing method demonstrated effectiveness in improving segmentation accuracy.
• The approach proved flexible and generalizable, enhancing various classic neural networks for tongue segmentation.

Conclusions:

• The developed method offers a significant advancement in precise tongue segmentation.
• The technique holds promise for improving the reliability of computer vision-based TCM diagnostic tools.
• The post-processing method's versatility suggests broad applicability in medical image analysis.