JMIR Publications

ABSTRACT

Background:

Deep learning models have attracted significant interest from health-care researchers during the last decades. There have been lots of studies that apply deep learning to medical applications and achieve promising performance. However, there are three limitations for existing works. (i) Most existing models are unable to interpret their results to clinicians. (ii) Existing models cannot incorporate complicated medical domain knowledge (e.g., a disease causes another disease). (iii) Most existing models lack visual exploration and interaction. Both the EHR dataset and the deep model results are complex and abstract, which impede clinicians to explore and communicate with the model directly.

Objective:

The objective of this study was to develop an interpretable and accurate risk prediction model, as well as an interactive clinical prediction system to support EHR data exploration, knowledge graph demonstration, and model interpretation.

Methods:

A domain knowledge guided recurrent neural network (RNN) model is proposed to predict clinical risks. The model takes medical event sequences as input, and incorporates medical domain knowledge by attending the medical knowledge graph. A global pooling operation and a fully connected layer are used output the clinical outcomes. The middle results and the parameters of the fully connected layer are helpful to identify which medical events cause clinical risks. Then DG-Viz is designed to support EHR data exploration, knowledge graph demonstration, and model interpretation.

Results:

We conduct both risk prediction experiments and a case study on a real-world dataset. 554 case patients with heart failure and 1662 control patients without heart failure are selected from the dataset. The experimental results show that the proposed DG-RNN outperforms the state-of-art approaches by about 1.5 percent. The case study demonstrates how our medical physician collaborator can effectively explore the data and interpret the prediction results using DG-Viz.

Conclusions:

In this work, we present DG-Viz, an interactive clinical prediction system, which brings together the power of deep learning (i.e., a domain knowledge guided RNN-based model) and visual analytics to predict clinical risks and visually interpret the EHR prediction results. Experimental results and a case study on heart failure risk prediction tasks demonstrate the effectiveness and usefulness of DG-Viz system. This study will pave the way for interactive, interpretable and accurate clinical risk predictions.