Research On Intelligent Diagnosis System Of Gastrointestinal Diseases Based On Tongue Images

In recent years,along with the improvement of people’s living standard,change of lifestyle and aging of population,the incidence of gastrointestinal diseases in China is increasing year by year,and the number of people suffering from gastrointestinal malignant tumors is gradually increasing.Therefore,the importance of early diagnosis of gastrointestinal diseases and early warning of cancer is highlighted.Traditional endoscopic and histopathological examinations have limitations in early diagnosis and screening of gastrointestinal diseases.Tongue diagnosis in traditional Chinese medicine can be a beneficial supplement to disease screening and diagnosis due to its non-invasive nature.Tongue diagnosis is an important part of "observation and diagnosis" in traditional Chinese medicine,which diagnoses diseases by observing the changes of the tongue.Tongue images can reflect the health status of the gastrointestinal tract,providing a new way for non-invasive auxiliary diagnosis of gastrointestinal diseases.Using artificial intelligence related technologies to achieve auxiliary diagnosis as well as non-invasive primary screening of diseases has become a current research hotspot.Based on artificial intelligence and computer image processing technology,the quantification and intelligence of tongue diagnosis can be achieved,which has high clinical application value for non-invasive auxiliary diagnosis of gastrointestinal diseases.It is an important driving force for the integration of Chinese and Western medicine and the modernization of tongue diagnosis in traditional Chinese medicine.This paper aims to construct an intelligent diagnosis system for gastrointestinal diseases based on tongue appearance.Traditional machine learning and deep learning methods are used to quantitatively analyze tongue image features,achieve accurate diagnosis of gastrointestinal diseases,and provide guidance for clinical applications.The main research contents are as follows:(1)Research on tongue image preprocessing methods.Aiming at the problems of poor segmentation results of traditional image segmentation algorithms for tongue images,the U-Net is used as the baseline segmentation model to achieve automatic segmentation of tongue images.Aiming at the problems of class-imbalance in the tongue image dataset,a hybrid sampling algorithm is proposed,which combines Borderline-SMOTE oversampling and fuzzy mean clustering undersampling.The problems of between-class and within-class imbalance of samples are considered,and the original data is sampled from the feature level to balance the class distribution.(2)Construction of a tongue diagnosis model for gastrointestinal diseases based on traditional machine learning methods.Aiming at the problems of inadequate image feature extraction and insufficient classification accuracy in traditional machine learning classification modeling,a classification model for gastrointestinal diseases based on information fusion method is constructed.This model linearly fuses the manual features and deep features of tongue images and constructs a decision fusion classifier using the Stacking method,which accomplishes both feature fusion and decision fusion.(3)Construction of a tongue diagnosis model for gastrointestinal diseases based on deep learning methods.Aiming at the problems of poor interpretability of deep learning disease diagnosis,a classification model for gastrointestinal diseases based on attention mechanism is constructed.This model consists of two parts:a feature extraction backbone network and a classification network based on the Slot attention module,where the spatial attention mechanism network is used to extract higher-order features and the self-attention mechanism network is used for pattern classification.The experimental results show that the fusion classification model can fully extract and fit the features of tongue images,which is an improvement and optimization of the traditional machine learning classification model.The deep learning model can effectively improve the diagnostic accuracy of gastrointestinal diseases and visually explain the diagnostic results by using attention mechanisms.The intelligent diagnosis system of gastroenterological diseases based on tongue images constructed in this paper has good classification accuracy and interpretability,and can provide assistance and guidance for clinical diagnosis and treatment.