Introduction

This repository contains the dataset and code utilized in our research on developing HM-TDF, a hard sample mining-based tongue diagnosis framework for fatty liver disease severity classification.

Paper Title

Hard Sample Mining-Based Tongue Diagnosis for Fatty Liver Disease Severity Classification

Method

We propose a Hard sample Mining-based Tongue Diagnosis Framework (HM-TDF) for fatty liver disease severity classification, which identifies hard-to-classify samples based on uncertainty and applies targeted learning to enhance classification performance, as shown in the following figure. We introduce a Multi-source Feature Fusion Kolmogorov-Arnold Network (MFF-KAN) and a three-stage training strategy, as shown in the following figure, to model the relationship between tongue images plus basic physiological indicators and FLD severity.

Released Dataset

The dataset, named Tongue-FLD, includes 5,717 samples: 3,690 with non-FLD, 1,512 with mild FLD, and 515 with moderate/severe FLD, resulting in an imbalance ratio of 7.17/2.94/1.00. Each sample includes a segmented tongue image, FLD severity annotation and eight physiological indicators. The physiological indicators include Gender, Age, Height, Waist circumference, hip circumference (Hipline), Weight, systolic blood pressure (SBP), and diastolic blood pressure (DBP). The data was obtained from a cohort study that received ethical approval. The participants were residents of Fuqing City, Fujian Province, China, aged 35 to 75 years. For each participant, a facial image with the tongue extended was captured using "TCM Diagnostic Devices," and basic physiological indicators were measured. Subsequently, participants underwent ultrasound examinations, and FLD severity was assessed according to the standard criteria established by the Fatty Liver Disease Study Group of the Chinese Liver Disease Association.

Usage

Python: 3.8.19

pip install -r requirements.txt
cd ./Tongue-FLD
cat Tongue_Images.tar.gz.* > Tongue_Images.tar.gz
tar xzf Tongue_Images.tar.gz
python random_rotate_images.py
python rotate_pre_train.py
python main.py

Cite this repository

If you find this code or dataset useful in your research, please consider citing us:
@inproceedings{HM-TDF,
title={Hard Sample Mining-Based Tongue Diagnosis for Fatty Liver Disease Severity Classification},
link={https://github.com/MLDMXM2017/HM-TDF}
}

Reference

https://github.com/KindXiaoming/pykan
https://github.com/Blealtan/efficient-kan
https://github.com/AntonioTepsich/Convolutional-KANs