Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Spectrotemporal Feature Extraction in EHG Signals and Tocograms for Enhanced Preterm Birth Prediction

Senith Jayakody, Kalana Jayasooriya, Sashini Liyanage, Roshan Godaliyadda, Parakrama Ekanayake, Chathura Rathnayake

TL;DR

It is demonstrated that combining denoising with domain-relevant features can yield highly accurate, robust, and clinically interpretable models, supporting the development of cost-effective and accessible PTB prediction tools, particularly in low-resource healthcare settings.

Abstract

Preterm birth (PTB), defined as delivery before 37 weeks of gestation, is a leading cause of neonatal mortality and long term health complications. Early detection is essential for enabling timely medical interventions. Electrohysterography (EHG) and tocography (TOCO) are promising non invasive tools for PTB prediction, but prior studies often suffer from class imbalance, improper oversampling, and reliance on features with limited physiological relevance. This work presents a machine learning (ML) pipeline incorporating robust preprocessing, physiologically grounded feature extraction, and rigorous evaluation. Features were extracted from EHG (and TOCO) signals using Mel frequency cepstral coefficients, statistical descriptors of wavelet coefficients, and peaks of the normalized power spectrum. Signal quality was enhanced via Karhunen Loeve Transform (KLT) denoising through eigenvalue based subspace decomposition. Multiple classifiers, including Logistic Regression, Support Vector Machines, Random Forest, Gradient Boosting, Multilayer Perceptron, and CatBoost, were evaluated on the TPEHGT dataset. The CatBoost classifier with KLT denoising achieved the highest performance on fixed interval segments of the TPEHGT dataset, reaching 97.28% accuracy and an AUC of 0.9988. Ablation studies confirmed the critical role of both KLT denoising and physiologically informed features. Comparative analysis showed that including TOCO signals did not substantially improve prediction over EHG alone, highlighting the sufficiency of EHG for PTB detection. These results demonstrate that combining denoising with domain-relevant features can yield highly accurate, robust, and clinically interpretable models, supporting the development of cost-effective and accessible PTB prediction tools, particularly in low-resource healthcare settings.

Spectrotemporal Feature Extraction in EHG Signals and Tocograms for Enhanced Preterm Birth Prediction

TL;DR

It is demonstrated that combining denoising with domain-relevant features can yield highly accurate, robust, and clinically interpretable models, supporting the development of cost-effective and accessible PTB prediction tools, particularly in low-resource healthcare settings.

Abstract

Preterm birth (PTB), defined as delivery before 37 weeks of gestation, is a leading cause of neonatal mortality and long term health complications. Early detection is essential for enabling timely medical interventions. Electrohysterography (EHG) and tocography (TOCO) are promising non invasive tools for PTB prediction, but prior studies often suffer from class imbalance, improper oversampling, and reliance on features with limited physiological relevance. This work presents a machine learning (ML) pipeline incorporating robust preprocessing, physiologically grounded feature extraction, and rigorous evaluation. Features were extracted from EHG (and TOCO) signals using Mel frequency cepstral coefficients, statistical descriptors of wavelet coefficients, and peaks of the normalized power spectrum. Signal quality was enhanced via Karhunen Loeve Transform (KLT) denoising through eigenvalue based subspace decomposition. Multiple classifiers, including Logistic Regression, Support Vector Machines, Random Forest, Gradient Boosting, Multilayer Perceptron, and CatBoost, were evaluated on the TPEHGT dataset. The CatBoost classifier with KLT denoising achieved the highest performance on fixed interval segments of the TPEHGT dataset, reaching 97.28% accuracy and an AUC of 0.9988. Ablation studies confirmed the critical role of both KLT denoising and physiologically informed features. Comparative analysis showed that including TOCO signals did not substantially improve prediction over EHG alone, highlighting the sufficiency of EHG for PTB detection. These results demonstrate that combining denoising with domain-relevant features can yield highly accurate, robust, and clinically interpretable models, supporting the development of cost-effective and accessible PTB prediction tools, particularly in low-resource healthcare settings.

Paper Structure

This paper contains 24 sections, 15 equations, 4 figures, 5 tables.

Table of Contents

  1. Introduction
  2. Materials and Methods
  3. Dataset
  4. Preprocessing
  5. Band-pass filtering
  6. Karhunen Loève Transform
  7. Segmentation and Sample Construction
  8. Contraction-dummy intervals
  9. Fixed 3-minute non overlapping intervals
  10. Feature Extraction
  11. Peak amplitude of the normalized power spectrum
  12. Mel coefficients extraction
  13. Wavelet coefficients extraction
  14. Model Selection
  15. Evaluation Protocol and Metrics
  16. ...and 9 more sections

Figures (4)

  • Figure 1: Proposed method for classifying PTBs using EHG recordings
  • Figure 2: Example power spectral density (PSD) of the EHG signal before (blue) and after (orange) applying a 4th order Butterworth band pass filter (0.08–5.0 Hz)
  • Figure 3: (a) Log-eigenvalue spectrum $\log(\lambda_i)$ in ascending order. (b) Consecutive log-eigenvalue differences $\Delta_i$. The 10% jump threshold ($\tau=0.1$) identifies the transition from noise-dominated components (left) to signal-dominated components (right) retained for reconstruction.
  • Figure 4: Model-wise classification AUCs on the 2012 TPEHG and 2018 TPEHGT datasets. Legend entries indicate segmentation type and KLT usage: C = contraction-based (2018), T = fixed 3-min windows with TOCO (2018), NT = fixed 3-min windows without TOCO (2012, 2018). Dashed lines = original signals, solid lines = KLT-denoised signals.