MUMBAI, India, July 13 -- Intellectual Property India has published a patent application (202641084274 A) filed by Kesavamoorthy R.; and Dayananda Sagar University on July 09, 2026, for Explainable Hybrid Deep Learning System And Method For Ecg-Based Arrhythmia Detection With Dual-Output Interpretability And Fault-Tolerant Decision Support.
Inventors include Sayed Faizan Qaser; Saransh Gupta; Dr. Santosh Reddy P; Dr. M. Shahina Parveen; Dr. Arunakumara B; Nishant Kumar Dube; Krishna Rathod; Siddharth Kumar; and Vivek Kumar.
The application for the patent was published on July 10, 2026, under issue no. 28/2026.
Abstract: The present invention discloses a computer-implemented system and method for explainable ECG-based arrhythmia detection with clinician-verifiable decision support. The system comprises an input module, a fault-tolerant validation module, a pre-processing module, a hybrid Convolutional Neural Network and Bidirectional Gated Recurrent Unit (CNN–BiGRU) inference engine trained under a composite loss combining a class-frequency-weighted focal loss with a rhythm-consistency regularisation term that penalises implausible beat-to-beat class transitions within a common RR-interval neighbourhood, a dual-output explainability head that jointly produces a waveform-region saliency overlay via gradient-weighted class activation on the ECG signal and a rule-grounded clinician- readable textual rationale grounded in measured clinical parameters including heart rate, PR interval, QT interval, QRS duration and an integrated cardiac risk score in the range 0 to 100, and a decision-support dashboard that renders the predicted class, model confidence, waveform saliency overlay, clinical parameters, textual rationale and recommendation. The invention is further embodied in a quantised edge-inference variant of the same architecture, sharing a common model artefact with a full-precision cloud variant, and configured to execute on an ARM-class mobile or near-patient processor with sub-100 millisecond per-beat inference latency. The invention addresses long-standing limitations of prior-art ECG classifiers by combining state-of-the-art accuracy, temporal plausibility of predictions, verifiable dual-modality explanations, direct clinical actionability, robust input handling and deployment flexibility into a single integrated decision-support system.
Disclaimer: Curated by HT Syndication.
