Facilitating the Development of Deep Learning Models with Visual Analytics for Electronic Health Records

Abstract

Electronic health record (EHR) data are widely used to perform early diagnoses and create treatment plans, which are key areas of research. We aimed to increase the efficiency of iteratively applying data-intensive technology and verifying the results for complex and big EHR data. We used a system entailing sequence mining, interpretable deep learning models, and visualization on data extracted from the MIMIC-IIIdatabase for a group of patients diagnosed with heart disease. The results of sequence mining corresponded to specific pathways of interest to medical staff and were used to select patient groups that underwent these pathways. An interactive Sankey diagram representing these pathways and a heat map visually representing the weight of each variable were developed for temporal and quantitative illustration. We applied the proposed system to predict unplanned cardiac surgery using clinical pathways determined by sequence pattern mining to select cardiac surgery from complex EHRs to label subject groups and deep learning models. The proposed system aids in the selection of pathway-based patient groups, simplification of labeling, and exploratory the interpretation of the modeling results. The proposed system can help medical staff explore various pathways that patients have undergone and further facilitate the testing of various clinical hypotheses using big data in the medical domain.

Keywords: deep learning models; electronic health records; visual analytics.

Figure 1

System architecture. CABG, coronary artery bypass grafting.

Figure 2

Architecture of the long short-term memory (LSTM) attention model. (a) The overall structure of the LSTM attention model and (b) the detailed process of the attention layer. Further, (c) shows the entire modeling process, including the data input and output.

Figure 2

Architecture of the long short-term memory (LSTM) attention model. (a) The overall structure of the LSTM attention model and (b) the detailed process of the attention layer. Further, (c) shows the entire modeling process, including the data input and output.

Figure 3

Example of a clinical pathway. (a) Example of a flowchart for sequenced events. (b) When an event is selected in the normal order, the link is highlighted in blue, and the selected card is displayed below. (c) When the event is selected in reverse order, it is highlighted in red, and the selected card is not displayed.

Figure 3

Example of a clinical pathway. (a) Example of a flowchart for sequenced events. (b) When an event is selected in the normal order, the link is highlighted in blue, and the selected card is displayed below. (c) When the event is selected in reverse order, it is highlighted in red, and the selected card is not displayed.

Figure 4

Attention heat maps. (a) Average attention for all patients in the case of cardiac surgery prediction in hospitals, (b) variableweights corresponding to specific patients, and (c) difference between the average weight for the patient group and variable weights corresponding to specific patients.

Figure 4

Attention heat maps. (a) Average attention for all patients in the case of cardiac surgery prediction in hospitals, (b) variableweights corresponding to specific patients, and (c) difference between the average weight for the patient group and variable weights corresponding to specific patients.

Figure 5

Visualization of the top 25 attention values of the in-hospital congestive heart failure diagnostic prediction model with a TreeMap. (a) Events are grouped by category. Selecting a category shows the name of the detailed events. (b) A detailed event screen that appears when the category “comorbidity” is selected.