Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

ConvexECG: Lightweight and Explainable Neural Networks for Personalized, Continuous Cardiac Monitoring

Rayan Ansari, John Cao, Sabyasachi Bandyopadhyay, Sanjiv M. Narayan, Albert J. Rogers, Mert Pilanci

TL;DR

Using data from 25 patients, it is demonstrated that ConvexECG achieves accuracy comparable to larger neural networks while significantly reducing computational overhead, highlighting its potential for real-time, low-resource monitoring applications.

Abstract

We present ConvexECG, an explainable and resource-efficient method for reconstructing six-lead electrocardiograms (ECG) from single-lead data, aimed at advancing personalized and continuous cardiac monitoring. ConvexECG leverages a convex reformulation of a two-layer ReLU neural network, enabling the potential for efficient training and deployment in resource constrained environments, while also having deterministic and explainable behavior. Using data from 25 patients, we demonstrate that ConvexECG achieves accuracy comparable to larger neural networks while significantly reducing computational overhead, highlighting its potential for real-time, low-resource monitoring applications.

ConvexECG: Lightweight and Explainable Neural Networks for Personalized, Continuous Cardiac Monitoring

TL;DR

Using data from 25 patients, it is demonstrated that ConvexECG achieves accuracy comparable to larger neural networks while significantly reducing computational overhead, highlighting its potential for real-time, low-resource monitoring applications.

Abstract

We present ConvexECG, an explainable and resource-efficient method for reconstructing six-lead electrocardiograms (ECG) from single-lead data, aimed at advancing personalized and continuous cardiac monitoring. ConvexECG leverages a convex reformulation of a two-layer ReLU neural network, enabling the potential for efficient training and deployment in resource constrained environments, while also having deterministic and explainable behavior. Using data from 25 patients, we demonstrate that ConvexECG achieves accuracy comparable to larger neural networks while significantly reducing computational overhead, highlighting its potential for real-time, low-resource monitoring applications.
Paper Structure (15 sections, 6 equations, 4 figures)

This paper contains 15 sections, 6 equations, 4 figures.

Table of Contents

  1. Introduction
  2. Preliminaries on Convex Neural Networks
  3. The Convex Formulation
  4. Model Interpretability
  5. Problem Formulation
  6. Approach
  7. Dataset Construction
  8. Data Preprocessing
  9. Baseline Models
  10. ECG Reconstruction Using Convex Neural Networks
  11. Experimental Results
  12. Reconstruction Results
  13. Variance and Convergence Analysis
  14. Analysis of Model Explainability
  15. Conclusions and Future Work

Figures (4)

  • Figure 1: Visualization of a Potential Multi-Model Reconstruction Paradigm. In the calibration stage, Leads I and II are recorded along with the ICM device signal to train the coupled ConvexECG model setup. In the deployment stage, the trained models operate on the ICM signal to reconstruct the 6-Lead ECG.
  • Figure 2: Comparison of model variance and performance. (A) Variance in predictions for ConvexECG vs. 2-layer (small) and 3-layer (medium) LSTMs. (B) Test set reconstructions for ConvexECG and LSTM initializations. (C) Train Mean Squared Error curves for 2-layer ReLU MLPs compared to ConvexECG.
  • Figure 3: Main panel shows 6-lead ECG reconstructions from the simulated ICM signal, with the training segment marked. Top right inset presents a box plot of Pearson correlation coefficients comparing reconstruction quality, while the bottom right inset illustrates model complexity by parameter count.
  • Figure 4: The learned $f_{\text{I}}$ plotted against input ICM values along with its training data, illustrating the link between the model’s behavior to specific datapoints.