Deep Learning-Based Method of Diagnosing Hyperlipidemia and Providing Diagnostic Markers Automatically

Liu, Yuliang; Zhang, Quan; Zhao, Geng; Liu, Guohua; Liu, Zhiang

A new method proposed can accurately and automatically diagnose hyperlipidemia and provide disease diagnostic markers at the same time. The visualization of the model diagnosis basis enhances the transparency of the black-box model, increases the interpretability of deep learning algorithm, and enhances the credibility of the model. The attention deep learning algorithm proposed in this paper realizes the providing diagnostic basis while diagnosing disease. This phenomenon proves that it has the potential to discover new diagnostic markers, and expands the application scope of the auxiliary diagnostic system. At the same time, the experimental results show that the algorithm also has the capability to learn the relationship between different physiological parameters, so it has a high generalization ability. Therefore, it can save medical resources, speed up the researching process of diagnostic markers to a certain extent, speed up the work efficiency of the hospital, and enhance the patient's medical experience. Increasing the explanatory power of the model can effectively increase the research on biomarkers. 34 The future work is still around to improve the performance of the auxiliary diagnostic system. In order to further improve the accuracy of the model, we will consider how to input more types of data into the model, such as patient history, etc. At the same time, in order to diagnose more kinds of diseases, we will collect more data to expand our existing data set. Because there are some complex diseases that require a joint judgment of multiple types of diagnostic information, we will study how to use cross-media diagnostic data as an input training model in the next step. Due to the limited data types, no new diagnostic markers are proposed in this model. Although the experiment confirmed that the diagnostic markers predicted by the model were the same as the gold standard, we will add more physiological parameter types and multiple diseases in the future work, with a view to finding more disease-related biomarkers. Not only in medicine but also from the perspective of engineering, we will further study the optimization methods of auxiliary diagnostic systems, such as the adjustment methods of hyperparameters. We will also further expand the sample data, consider more factors that may influence the diagnosis of the disease such as different races, diverse age groups et al to further enhance the reliability of the model.