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

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

In a limited range, we found similar work. Edward Choi and his colleagues used recurrent neural networks to process electronic health records of varying lengths for early diagnosis of heart failure, reaching 88.3% of AUC. 6 Michael A. Schwemmer et al used a deep neural network decoding framework to classify intracortical recording, reaching 93.78% ACC. 10 Oliver Faust et al used LSTM neural network to process RR interval signals for automatic diagnosis of atrial fibrillation. 36 All the work had achieved better performance in the test set. Although EHR data are widely used in the research of auxiliary diagnostic system, there is no unified standard to evaluate the quality of EHR data at present. The EHR data include artificial description, which limits the credibility of EHR data, which is also one of the important factors limiting the further improvement of model performance. Moreover, because EHR data does not have a uniform format, it is necessary to extract features manually before data are utilized, which not only causes the loss of original information but also increases labor costs. In the training process of this model, physiological parameters with standardized criteria were applied to the training of the model, and there was no manual description process. At the same time, the proposed model does not need to manually extract features, so that the model can obtain more potentially useful information, thus improving the performance of the model and increasing the reliability of the model. In addition, the explanation of disease mechanism and biomarker should be added. Only the improvement of diagnostic accuracy can be used to prove that the improvement of medical diagnostic technology is not very comprehensive. The accuracy of diagnosis is difficult to represent the level of comprehensive diagnosis.