A Hybrid Quantum-inspired Framework for Heart Disease Prediction: Integrating Quantum Genetic Algorithm and Quantum Particle Swarm Optimization With Support Vector Machine and Self-learning Capabilities
Year: 2026
Type: article
Abstract: Cardiovascular disease accounts for more deaths annually than any other condition; the WHO recorded 17.9 million fatalities in 2022 alone. Early, automated screening from routine clinical measurements is therefore a priority with direct mortality implications. Three persistent shortcomings limit current approaches: feature selectors routinely retain correlated, redundant inputs that weaken classification; SVM hyperparameter search still depends heavily on grid search, which scales poorly as resolution increases; and trained models are seldom refreshed, degrading silently as patient populations shift. This paper introduces QI-HeartNet, an end-to-end cardiac prediction pipeline that addresses all three within a unified framework. Feature selection is performed by a Quantum-Inspired Genetic Algorithm (QGA), where each candidate feature is encoded as a qubit amplitude pair rather than a fixed binary flag; chromosomes are initialised at equal superposition and updated through rotation gates, sustaining search diversity into late generations a property that conventional binary-string GAs cannot preserve. SVM hyperparameters C and gamma are optimised by Quantum Particle Swarm Optimisation (QPSO), whose delta-potential-well position update allows particles to escape local traps that pin down standard PSO. Both quantum-inspired paths are benchmarked against classical counterparts on the 918-sample UCI Heart Disease Combined Dataset under stratified 10-fold cross-validation. On the 184-patient held-out test set, QGA-QPSO-QSVM achieved 85.87% accuracy, 91.18% sensitivity, 91.74% AUC-ROC, and a Diagnostic Odds Ratio of 39.51 outperforming the classical pair on every primary clinical metric. A PSI-monitored online retraining component reduced false negatives by 18% relative to the static model in a controlled retrospective simulation.
Authors: Kiran Dethe, S.M. Turkane
Institution: Savitribai Phule Pune University
Subfield: Health Information Management
Field: Health Professions
Domain: Health Sciences
Open Access status: diamond
OpenAlex ID: W7166357668