A Visual Analytics Approach to Exploring the Feature and Label Space Based on Semi-structured Electronic Medical Records



He Wang, Yang Ouyang, Quan Li

 Room: 104

2023-10-22T03:00:00ZGMT-0600Change your timezone on the schedule page
2023-10-22T03:00:00Z

Exemplar figure, described by caption below — System Interface: (A) The interface panel showcases statistical information concerning the dataset and model. (B) Label Identification View aids clinicians in recognizing potential labels from unstructured diagnostic texts. (C) Feature Exploration View empowers clinicians to choose features by evaluating feature distribution and significance. (D) ML Modeling and Interpretation View reveals the distribution of features among diverse groups and the importance of features contributing to distinct categories.

Fast forward

Keywords

Human-centered computing—Visualization—Visualization techniques

Abstract

Electronic health records (EHRs), serving as patient-centered repositories for medical data, offer the opportunity for researchers to uncover concealed patterns using machine learning (ML). However, in real-world medical settings, clinicians often face the task of selecting pertinent feature dimensions from a range of potential medical metrics and then deducing potential labels from vague diagnostic descriptions, prior to the modeling phase. This complexity presents challenges in obtaining reliable training/testing data and conducting thorough analysis. Consequently, these hurdles hinder the practical application of ML for automated modeling and comprehensible interpretation of influencing factors. To tackle these challenges, we introduce a visual analytics approach designed to navigate the feature and label space within EHRs, while also streamlining the modeling process through automated ML algorithms and techniques for improved interpretability.