Weakly Supervised Object Detection for Automatic Tooth-marked Tongue Recognition

TL;DR

This paper introduces a fully automated weakly supervised method utilizing Vision Transformer and multiple instance learning for accurate detection and recognition of tooth-marked tongues in clinical images, improving objectivity and diagnostic precision in TCM.

Contribution

It presents a novel end-to-end weakly supervised approach combining ViT and MIL for tongue region detection and tooth-marked tongue recognition, with high accuracy and clinical relevance.

Findings

High accuracy in tooth-marked tongue classification

Effective localization of tooth-marked regions

Automated process improves diagnostic objectivity

Abstract

Tongue diagnosis in Traditional Chinese Medicine (TCM) is a crucial diagnostic method that can reflect an individual's health status. Traditional methods for identifying tooth-marked tongues are subjective and inconsistent because they rely on practitioner experience. We propose a novel fully automated Weakly Supervised method using Vision transformer and Multiple instance learning WSVM for tongue extraction and tooth-marked tongue recognition. Our approach first accurately detects and extracts the tongue region from clinical images, removing any irrelevant background information. Then, we implement an end-to-end weakly supervised object detection method. We utilize Vision Transformer (ViT) to process tongue images in patches and employ multiple instance loss to identify tooth-marked regions with only image-level annotations. WSVM achieves high accuracy in tooth-marked tongue classification, and visualization experiments demonstrate its effectiveness in pinpointing these regions. This automated approach enhances the objectivity and accuracy of tooth-marked tongue diagnosis. It provides significant clinical value by assisting TCM practitioners in making precise diagnoses and treatment recommendations. Code is available at https://github.com/yc-zh/WSVM.