Topological Analysis for Sleep Apnea Detection: Project Proposal

Dylan LI, Yeap Yun Ting

Keywords: sleep apnea, classification, ECG, topological data analysis, persistent homology

Abstract:

This paper presents a research proposal for a topological data analysis (TDA) framework for sleep apnea detection from single-lead ECG signals. Our approach aims to leverage persistent homology to capture intrinsic geometric structures that are often overlooked by standard machine learning methods.

We plan to evaluate the proposed method on the PhysioNet Apnea-ECG dataset and assess whether TDA-based representations can achieve competitive performance while relying solely on cardiac data. In particular, we aim to investigate whether topological features provide complementary information to conventional approaches and enable reliable ECG-only detection.

This work seeks to explore the potential of TDA as a scalable and interpretable alternative for automated sleep apnea detection.

Submission Number: 2