Analyzing EHR data using DL Methods Sample Clauses

Analyzing EHR data using DL Methods. To the best of our knowledge, although deep learning methods have been widely applied in many fields, it is in recent years that researchers start to analyze EHR data using DL methods. We review two related works here. The work by Cheng et al. (2016+) directly adopt the regular convolutional neural network model. To accommodate the temporal features of EHR data, their model consists of four layers - a input layer with the temporal matrix of EHR data, a convo- lutional layer with different window sizes, a pooling layer with additional normaliza- tion function, and a fully connected layer using softmax function. In addition, they proposed different fusing structure to fully utilize the temporal pattern in EHR data. Figure 1.1 demonstrates the four-layered neural network proposed in this paper. They examine the methods using a Congestive Heart Failure (CHD) cohort (1127 cases and 3850 controls) and a Chronic Obstructive Pulmonary Disease (COPD) cohort (477 cases and 2385 controls). And the proposed models can achieve a best AUC of around