Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

TopSeg: A Multi-Scale Topological Framework for Data-Efficient Heart Sound Segmentation

Peihong Zhang, Zhixin Li, Yuxuan Liu, Rui Sang, Yiqiang Cai, Yizhou Tan, Shengchen Li

TL;DR

TopSeg, a topological representation-centric framework that encodes PCG dynamics with multi-scale topological features and decodes them using a lightweight temporal convolutional network (TCN) with an order- and duration-constrained inference step, indicates that topology-aware representations provide a strong inductive bias for data-efficient, cross-dataset PCG segmentation, supporting practical use when labeled data are limited.

Abstract

Deep learning approaches for heart-sound (PCG) segmentation built on time--frequency features can be accurate but often rely on large expert-labeled datasets, limiting robustness and deployment. We present TopSeg, a topological representation-centric framework that encodes PCG dynamics with multi-scale topological features and decodes them using a lightweight temporal convolutional network (TCN) with an order- and duration-constrained inference step. To evaluate data efficiency and generalization, we train exclusively on PhysioNet 2016 dataset with subject-level subsampling and perform external validation on CirCor dataset. Under matched-capacity decoders, the topological features consistently outperform spectrogram and envelope inputs, with the largest margins at low data budgets; as a full system, TopSeg surpasses representative end-to-end baselines trained on their native inputs under the same budgets while remaining competitive at full data. Ablations at 10% training confirm that all scales contribute and that combining H_0 and H_1 yields more reliable S1/S2 localization and boundary stability. These results indicate that topology-aware representations provide a strong inductive bias for data-efficient, cross-dataset PCG segmentation, supporting practical use when labeled data are limited.

TopSeg: A Multi-Scale Topological Framework for Data-Efficient Heart Sound Segmentation

TL;DR

TopSeg, a topological representation-centric framework that encodes PCG dynamics with multi-scale topological features and decodes them using a lightweight temporal convolutional network (TCN) with an order- and duration-constrained inference step, indicates that topology-aware representations provide a strong inductive bias for data-efficient, cross-dataset PCG segmentation, supporting practical use when labeled data are limited.

Abstract

Deep learning approaches for heart-sound (PCG) segmentation built on time--frequency features can be accurate but often rely on large expert-labeled datasets, limiting robustness and deployment. We present TopSeg, a topological representation-centric framework that encodes PCG dynamics with multi-scale topological features and decodes them using a lightweight temporal convolutional network (TCN) with an order- and duration-constrained inference step. To evaluate data efficiency and generalization, we train exclusively on PhysioNet 2016 dataset with subject-level subsampling and perform external validation on CirCor dataset. Under matched-capacity decoders, the topological features consistently outperform spectrogram and envelope inputs, with the largest margins at low data budgets; as a full system, TopSeg surpasses representative end-to-end baselines trained on their native inputs under the same budgets while remaining competitive at full data. Ablations at 10% training confirm that all scales contribute and that combining H_0 and H_1 yields more reliable S1/S2 localization and boundary stability. These results indicate that topology-aware representations provide a strong inductive bias for data-efficient, cross-dataset PCG segmentation, supporting practical use when labeled data are limited.
Paper Structure (12 sections, 7 equations, 3 figures, 4 tables)

This paper contains 12 sections, 7 equations, 3 figures, 4 tables.

Table of Contents

  1. Introduction
  2. Topological Data Analysis
  3. Proposed TopSeg Framework
  4. Multi-scale Topological Feature Extraction
  5. Topology-Only Constraint-Aware Refinement
  6. Experiment and Result
  7. Dataset
  8. Data Preprocessing
  9. Comparison Models
  10. Results under Data-Efficient Training
  11. Multi-Scale Ablation
  12. Conclusion

Figures (3)

  • Figure 1: PCG signal segmented into S1, systole, S2, and diastole.
  • Figure 2: TDA overview. (a) Time-delay embedding of a heart-sound segment; repeating cardiac cycles induce recurrent structures in the reconstructed phase space. (b) Filtration of simplicial complexes indexed by the scale parameter $\varepsilon$ (illustrated with an $\alpha$-complex), from which PH collects birth--death pairs for $H_0/H_1$. (c) Persistence landscapes $\{\lambda_k\}$ obtained from the diagram provide stable, fixed-length descriptors amenable to learning.
  • Figure 3: Illustration of the TopSeg framework.