Application of U-Net with Global Convolution Network Module in Computer-Aided Tongue Diagnosis

Area of Science:

• Artificial Intelligence
• Medical Imaging
• Computer Vision

Background:

• Intelligent manufacturing supports smart medical services, with a focus on AI for medical diagnosis.
• Tongue crack analysis is crucial for diagnosing diseases and Traditional Chinese Medicine (TCM) syndromes.
• Existing computer tongue diagnosis systems lack in-depth research on fissured tongues.

Purpose of the Study:

• To develop an improved deep learning model for accurate semantic segmentation of fissured tongue images.
• To enhance the feature extraction capabilities of existing U-Net architectures for tongue image analysis.
• To contribute to the development of more accurate computer-aided tongue diagnosis systems.

Main Methods:

• Proposed an improved U-Net network incorporating a Global Convolution Network module.
• Integrated the module into the encoder of the U-Net architecture to enhance semantic feature extraction.
• Validated the method using a dedicated fissured tongue image dataset.

Main Results:

• The improved U-Net network demonstrated superior segmentation performance compared to standard models.
• Achieved higher segmentation accuracy on the fissured tongue image dataset.
• The model effectively extracts abstract, high-level semantic features crucial for diagnosis.

Conclusions:

• The proposed improved U-Net network is effective for semantic segmentation of fissured tongues.
• This advancement can significantly enhance the accuracy of computer-aided tongue diagnosis systems.
• Further research in this area can improve patient safety and clinical efficiency.