Copyright © 2026 Authors retain the copyright of this article. This article is an open access article distributed under the Creative Commons Attribution License which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.

@article{199385,
        author = {Nikhilesh Kancherla and Chavan Rupesh Sharan and Cooly Siddarttha and Ajay Bhaskar Reddy Satti and Dr.Kirithiga Nandini},
        title = {Performance and Interpretability Trade-offs in Mixed-Precision and Quantized ECG Arrhythmia Classifiers for Edge Deployment},
        journal = {International Journal of Innovative Research in Technology},
        year = {2026},
        volume = {12},
        number = {11},
        pages = {13038-13044},
        issn = {2349-6002},
        url = {https://ijirt.org/article?manuscript=199385},
        abstract = {Continuous ECG monitoring is becoming more and more important for finding arrhythmias early, especially in wearable and edge-based healthcare systems where memory, power, and latency are all limited. This project looks at how different deep learning models act when they are optimized for deployment using mixed precision and quantization, while still making sure that their predictions are accurate. This work uses the MIT-BIH Arrhythmia Database to compare the 1D-CNN, ResNet1D, BiLSTM, and Transformer1D models in FP32, FP16/BF16, and INT8 settings using Post-Training Quantization and Quantization-Aware Training. The project looks at both classification performance and interpretability. It uses Saliency, SHAP, and LIME to see if compressed models still focus on important ECG areas like the QRS complex and strange beat patterns. The study is set up as a full pipeline with preprocessing, training, optimization, ONNX export, and explanation analysis. This makes it good for real-time edge deployment. In general, the work shows that model compression can greatly improve efficiency. However, the best practical balance is reached when accuracy, latency, and explanation stability are all taken into account at the same time.},
        keywords = {ECG arrhythmia classification, edge deployment, mixed precision, quantization, explainable AI, MIT-BIH arrhythmia database, interpretability drift.},
        month = {April},
        }