Supervised Laplacian Eigenmaps with Applications in Clinical Diagnostics for Pediatric Cardiology

TL;DR

This paper introduces a supervised Laplacian eigenmaps method that enhances machine learning models with low-dimensional textual data representations, improving diagnostic accuracy in pediatric cardiology from electronic health records.

Contribution

The paper presents a novel supervised Laplacian eigenmaps approach that effectively integrates textual and vector data for improved clinical diagnostics.

Findings

Supervised Laplacian eigenmaps outperformed other methods in AUC and MCC.

Achieved 8.16% increase in AUC over baseline without text.

Enhanced diagnostic prediction accuracy in pediatric cardiology.

Abstract

Electronic health records contain rich textual data which possess critical predictive information for machine-learning based diagnostic aids. However many traditional machine learning methods fail to simultaneously integrate both vector space data and text. We present a supervised method using Laplacian eigenmaps to augment existing machine-learning methods with low-dimensional representations of textual predictors which preserve the local similarities. The proposed implementation performs alternating optimization using gradient descent. For the evaluation we applied our method to over 2,000 patient records from a large single-center pediatric cardiology practice to predict if patients were diagnosed with cardiac disease. Our method was compared with latent semantic indexing, latent Dirichlet allocation, and local Fisher discriminant analysis. The results were assessed using AUC, MCC, specificity, and sensitivity. Results indicate supervised Laplacian eigenmaps was the highest performing method in our study, achieving 0.782 and 0.374 for AUC and MCC respectively. SLE showed an increase in 8.16% in AUC and 20.6% in MCC over the baseline which excluded textual data and a 2.69% and 5.35% increase in AUC and MCC respectively over unsupervised Laplacian eigenmaps. This method allows many existing machine learning predictors to effectively and efficiently utilize the potential of textual predictors.