Paperdoi:10.1016/j.bios.2026.119264
SF2O-SKDCN: Selective kernel attention enabled deep learning framework for cardiovascular disease detection using photoplethysmography signals
Biosensors and Bioelectronics 316:119264 (online 2026-10-09; issue Jan 2027)
Licence: Elsevier subscription (paywalled)
TL;DR
A deep CNN with "Selective Kernel Shuffle" attention classifies cardiovascular disease from segmented PPG, with a heart-rate feature descriptor, and its hyperparameters are tuned by a new "Sail Tuna Fish Foraging Optimization" (SF2O) metaheuristic. On MIMIC II with a 90 % training split it reports 98.54 % accuracy, 99.07 % recall, 98.05 % PPV and 99.05 % NPV.
- Form factor: clinical_finger
Abstract
Cardiovascular Disease (CVD) is a lethal disease that needs early intervention to prevent further harm and eventually safeguard patients' lives. Therefore, CVD detection is critical for effective disease management. Even though practical findings are made using traditional detection methods, they mostly struggle with overfitting risks, generalizability issues, and computational overhead challenges during the detection process. Therefore, the research proposes Sail Tuna Fish Foraging Optimized-Selective Kernel attention-enabled Deep Convolutional Neural Network (SF2O-SKDCN) that utilizes photoplethysmography (PPG) signals, which are extensively used in clinical applications as a screening tool, and the heart rate feature descriptor is introduced to extract the crucial feature from the segmented signal. Further, the Selective Kernel Shuffle (SKS) attention enables the model to focus only on the predominant CVD features, thereby enhancing the interpretability characteristics. In addition, actively tuning the model hyperparameters using the Sail Tuna Fish Foraging Optimization (SF2O) algorithm prevents local convergence and vanishing gradient issues. Experimenting with the MIMIC II dataset, the model yields better results by achieving 98.54% accuracy, 99.07% recall, 98.05% Positive predictive value, and 99.05% Negative predictive value for 90% training.
Verbatim from Europe PMC (opens in a new tab) · Abstract via Europe PMC; © the authors/publisher
Results on ChatPPG leaderboards
All leaderboardsNo results from this paper on ChatPPG leaderboards yet. Browse the leaderboards.
Tasks & signals
Tasks
Signals
Keywords
How it works
Cite this paper
APA
Chavan, P. M., Joshi, K. R., & Kamathe, R. S. (2027). SF2O-SKDCN: Selective kernel attention enabled deep learning framework for cardiovascular disease detection using photoplethysmography signals. Biosensors and Bioelectronics. https://doi.org/10.1016/j.bios.2026.119264
BibTeX
@article{chavan2027sf2oskdcn,
title = {{SF2O-SKDCN: Selective kernel attention enabled deep learning framework for cardiovascular disease detection using photoplethysmography signals}},
author = {Chavan, Pratima M. and Joshi, Kalyani R. and Kamathe, Rupali S.},
year = {2027},
journal = {Biosensors and Bioelectronics},
doi = {10.1016/j.bios.2026.119264}
}Built from the hub record. Check it against the publisher before submitting.
Added to the ChatPPG research hub
