Key result
Novel hybrid model achieves ~99% accuracy in ECG arrhythmia classification across standard databases.
Why the study?
Accurate ECG arrhythmia detection for early detection and wireless cardiac monitoring remains difficult due to complex temporal patterns, morphological differences, noise, individual variations, and a highly imbalanced class problem.
Does the hybrid SLNN-TDA-HQHS machine learning framework improve automated ECG arrhythmia classification accuracy compared to existing state-of-the-art methods in public arrhythmia datasets?
Does the hybrid SLNN-TDA-HQHS machine learning framework improve automated ECG arrhythmia classification accuracy compared to existing state-of-the-art methods in public arrhythmia datasets?
The hybrid SLNN-TDA-HQHS computational framework achieved 99.26% and 99.00% arrhythmia classification accuracy across two benchmark ECG datasets, demonstrating potential for real-time personalized cardiac monitoring and clinical decision support.
A precise ECG arrhythmia classification is the prerequisite for early detecting and wireless cardiac monitoring. Nonetheless, accurate detection is still difficult owing to complex temporal patterns, morphological differences, noises, individual variations, as well as a highly imbalanced class problem. In this study, we propose a hybrid approach, SLNN-TDA-HQHS, which combines Spiking Liquid Neural Networks (SLNN), Topological Data Analysis (TDA) and Quantum-Inspired Hippopotamus Harmony Search (HQHS) optimization. SLNN models the nonlinear temporal dependencies and multi-scale properties of the original ECG signals, TDA provides noise-sensitive topological features based on persistent homology. An attention based fusion module is also proposed to combine temporal and topological representations, with the aim to further enhance feature discriminability. HQHS performs joint optimization of feature selection and model parameters, which is a more efficient search strategy in high-dimensional space. For better interpretability, a prototype-driven contrastive explanation module locates prototypical patterns in support and opposition of classification decisions. Online transfer learning also allows customization based on patient-specific ECG traits. The proposed methodology was tested on two datasets that is the UCI Arrhythmia Dataset and the INCART 2-Lead Arrhythmia Database with success rates of 99.26% and 99.00% respectively. The results illustrate superior noise and class imbalance robustness and functional potential of the proposed framework for real-time personalized cardiac monitoring and clinical decision support applications.
No takes yet. Share an insight, caveat, or question.
Mhetre et al. (2026) studied Cardiac arrhythmia (n=527). SLNN-TDA-HQHS hybrid framework (Spiking Liquid Neural Networks, Topological Data Analysis, and Quantum-Inspired Hippopotamus Harmony Search) vs. State-of-the-art machine learning and deep learning models was evaluated on ECG arrhythmia classification accuracy on the UCI Arrhythmia Dataset. The SLNN-TDA-HQHS hybrid framework achieved ECG arrhythmia classification accuracies of 99.26% on the UCI Arrhythmia dataset and 99.00% on the INCART 2-Lead Arrhythmia database.
Synapse has enriched 5 closely related papers on similar clinical questions. Consider them for comparative context: