Research on Chronic Kidney Disease Staging Prediction and Prescription Recommendation Model Based on Multimodal Data Fusion

Chronic kidney disease is common and prevalent in China. Traditional Western medicine mainly relies on numerical data for disease condition analysis and staging inference. However, due to the insidious nature of chronic diseases, data from a single modality often fails to reflect the full picture of the disease, and there is a lack of the ability to comprehensively evaluate the disease development. In addition, due to the serious phenomenon of irregular medication globally, the shortage of professional pharmacists in primary medical institutions in China, and the difficulty in inheriting traditional Chinese medicine (TCM) experience, there are defects in traditional prescription recommendation methods. This paper aims to propose a study on a prediction model for the staging of chronic kidney disease and a prescription recommendation model based on multi-modal data fusion. Firstly, numerical clinical data from Western medicine, text-based diagnostic description data from TCM, and TCM tongue image data are selected. An improved BP neural network model is used to process the numerical clinical blood test data from Western medicine, a convolutional neural network model is applied to handle the TCM tongue image data, and an XGBoost model is adopted to deal with the TCM text data. Finally, the image data is converted into text data, and multiple types of text data are weighted and fused with numerical data to achieve the staging prediction of chronic kidney disease. Therefore, based on the data of the chronic kidney disease (CKD) staging results, a TCM prescription recommendation model is studied. The BERT model based on the Transformer architecture is used to process the symptom and drug texts, and multiple machine learning algorithms are employed to construct the recommendation model.