Research On The Application Of Multi-Label Classification Based On Deep Learning In The Auxiliary Diagnosis Of Asthma In TCM

Currently,as the modernization process of Traditional Chinese Medicine(TCM)accelerates and the quality of medical services continuously improves,TCM assisted diagnosis has become an inevitable trend and direction of TCM development.TCM assisted diagnosis technology combines TCM diagnostic thinking with artificial intelligence technology to intelligently analyze and process patient information,which can assist TCM doctors in more accurately diagnosing and treating diseases,thereby improving the efficiency and quality of medical services.The core of traditional Chinese medicine theory lies in syndrome differentiation and treatment,which serves as the fundamental principle for disease treatment and diagnosis in TCM.Electronic medical records of TCM contain a vast amount of clinical diagnostic and treatment experiences from TCM practitioners,as well as medical diagnosis data from patients.Exploring the rich experiences in syndrome differentiation and treatment embedded in these electronic medical records of TCM is of significant practical importance.The syndrome reflected the nature of a specific stage in the disease progression and the specific internal and external environment in which the individual patient is present through the information of visiting,hearing,consultation and cutting,which provides a basis for dialectical treatment.Symptoms can be divided into causes,pathogenesis,and locations,and etiology,pathogenesis,and disease locations each contain multiple categories,and the categories can be combined with each other.When analyzing and predicting TCM syndromes,if only seen as simple text classification problems,the stable syndromes appearing in a certain stage of the disease process are classified and causality is simply inferred,ignoring the parallel relationships between syndrome conclusions.This leads to the neglect of important clinical information and affects the effectiveness of TCM diagnosis and treatment.TCM texts contain a large number of professional terms in the field of TCM,and the correlation between contextual words is strong,and it contains ancient texts with rigorous expression and dialectical thinking,which is highly professional,and it is difficult to carry out textual representation and semantic understanding.To address these issues,this paper conducts experimental research using electronic medical records of traditional Chinese medicine for asthma patients provided by Qihuang National College at Jiangxi University of Traditional Chinese Medicine as a data sample.The main work of this article is as follows:(1)The field of TCM contains a large amount of unlabeled data,and the semantic information of texts extracted by traditional multi-label classification models is not complete.Aiming to address this issue,a multi-label classification model for Chinese medical text is proposed,which combines semantic filtering using ALBERT and the Text CNN neural network.Firstly,a specific domain task is conducted for self-training,in which the unannotated texts in the asthma domain are used as input for pre-training of the multi-label classification model with ALBERT.Secondly,the multi-layers of Transform in ALBERT are used to dynamically vectorize the annotated data and generate efficient text vectors based on semantic selection from the best encoding layer.Finally,Text CNN is introduced to establish a multi-label classifier and extract semantic information features from text vectors at different levels.To validate the effectiveness of the method,tests were conducted on a traditional Chinese medicine dataset,and the results show improved classification accuracy of the model.(2)Traditional multi-label classification models often fail to fully consider the complex correlations between adjacent labels in current text.To address this issue,we propose a multi-label classification method for Traditional Chinese Medicine(TCM)text,which is based on a local attention Seq2 Seq approach.Firstly,the ALBERT model is used to extract the dynamic semantic vector of the text.Then the coding layer composed of multiple layers of Bi-LSTM is used to extract the semantic relationship between texts;Finally,the local attention of multiple layers of LSTMs is used in the decoding layer to highlight the mutual influence between adjacent labels in the text sequence to predict multi-label sequences.The effectiveness of the proposed method is verified on a TCM dataset,and the experimental results demonstrate that the algorithm can effectively capture the correlations between labels,making it suitable for classifying and predicting TCM texts.(3)Building upon the aforementioned research and considering the practical demands of clinical auxiliary diagnosis applications,this paper adopts PYTHON programming language and related development tools to design and develop an auxiliary diagnosis system for TCM asthma on the Web side based on the Flask framework.