Abstract
This paper proposed a feature selection method combined with multi-time-scales analysis and heart rate variability (HRV) analysis for middle and early diagnosis of congestive heart failure (CHF). In previous studies regarding the diagnosis of CHF, researchers have tended to increase the variety of HRV features by searching for new ones or to use different machine learning algorithms to optimize the classification of CHF and normal sinus rhythms subject (NSR). In fact, the full utilization of traditional HRV features can also improve classification accuracy. The proposed method constructs a multi-time-scales feature matrix according to traditional HRV features that exhibit good stability in multiple time-scales and differences in different time-scales. The multi-scales features yield better performance than the traditional single-time-scales features when the features are fed into a support vector machine (SVM) classifier, and the results of the SVM classifier exhibit a sensitivity, a specificity, and an accuracy of 99.52%, 100.00%, and 99.83%, respectively. These results indicate that the proposed feature selection method can effectively reduce redundant features and computational load when used for automatic diagnosis of CHF.
Introduction
Congestive Heart Failure (CHF) is a chronic condition where the heart's pumping efficiency is compromised, necessitating early and accurate diagnosis to improve patient outcomes. Heart Rate Variability (HRV)—the variation in time between consecutive heartbeats—has emerged as a powerful non-invasive diagnostic tool. However, current research often focuses on either discovering complex new mathematical biomarkers or applying sophisticated machine learning models to short, isolated data snapshots.
This study introduces a novel approach by rethinking the application of traditional HRV metrics. By analyzing these standard features across multiple time scales—ranging from 5 minutes to 10 hours—the researchers demonstrate that near-perfect diagnostic accuracy can be achieved without the need for complex new features.
Methodology: Multi-Time-Scale Analysis
The core of the proposed method is a feature selection scheme that prioritizes both classification significance and stability across various temporal windows. The workflow is divided into four primary stages:
- Segmentation: Long-term RR interval time series are segmented into seven distinct time scales: 5 min, 10 min, 30 min, 1 hour, 2 hours, 5 hours, and 10 hours.
- Feature Extraction: Nine traditional HRV features are calculated across three domains: Time (MEAN, SDNN, RMSSD), Frequency (LFn, HFn, Ratio-LH), and Nonlinear (VAI, VLI, SampEn).
- Feature Selection: Features are selected based on their ability to differentiate between healthy (NSR) and CHF subjects consistently across all time scales. The final selection includes MEAN, RMSSD, LFn, HFn, and VAI.
- Classification: A Support Vector Machine (SVM) classifier, utilizing leave-one-out cross-validation (LOOCV), processes the multi-time-scale feature matrix to perform the diagnosis.
Results and Insights
The study utilized a dataset of 83 subjects, including 54 healthy individuals (Normal Sinus Rhythm) and 29 patients with CHF. The primary finding is that the stability of a feature across time scales serves as a reliable proxy for its diagnostic robustness.
By filtering for features that remain significant from short-term (5-minute) to long-term (10-hour) observations, the researchers effectively removed the "noise" that often compromises single-scale analyses. As shown in the comparative analysis of various HRV metrics, certain features like MEAN and RMSSD maintain a clear separation between the two groups across all observed time windows, whereas others may fluctuate.
The multi-time-scale SVM classifier achieved exceptional performance:
- Sensitivity: 99.52%
- Specificity: 100.00%
- Accuracy: 99.83%
These results significantly outperform traditional single-time-scale methods, proving that the temporal "depth" of the data is a critical factor in diagnostic accuracy.
Discussion and Future Work
While the results are promising, several considerations remain for clinical implementation. The study's reliance on a relatively small dataset (83 samples) suggests a need for validation against larger, more diverse populations. Furthermore, the requirement for 10-hour data windows positions this method as a tool for long-term monitoring or routine check-ups rather than immediate emergency triage.
Future research may explore applying this multi-scale framework to other cardiac conditions, potentially establishing a universal "stability-based" feature selection method for EKG analysis. By maximizing the utility of traditional metrics, this approach offers a path toward more transparent and computationally efficient automated diagnostics.
Summary updated