KACQ-DCNN: Uncertainty-Aware Interpretable Kolmogorov-Arnold Classical-Quantum Dual-Channel Neural Network for Heart Disease Detection

Heart failure remains a leading cause of global mortality, demanding improved diagnostic strategies that overcome limitations in classical machine learning such as high-dimensional data challenges, class imbalances, poor feature representations, and lack of interpretability. We propose KACQ-DCNN (Kolmogorov-Arnold Classical-Quantum Dual-Channel Neural Network), a novel hybrid architecture that replaces traditional multilayer perceptrons with Kolmogorov-Arnold Networks (KANs) featuring learnable univariate activation functions. Our 4-qubit, 1-layer KACQ-DCNN model outperforms 37 benchmark models—including 16 classical and 12 quantum neural networks—achieving an accuracy of 92.03% with macro-average precision, recall, and F1 scores of 92.00%, and a ROC-AUC of 94.77%, surpassing competitors by significant margins validated through paired t-tests (significance threshold of 0.0056 after Bonferroni correction). Ablation studies demonstrate the synergistic effect of classical-quantum integration, improving performance by approximately 2% over MLP variants. The framework integrates LIME and SHAP explainability techniques for enhanced feature interpretability and employs conformal prediction for robust uncertainty quantification, making it not only highly accurate but also clinically trustworthy for cardiovascular diagnostics.

Recommended citation: Jahin, M. A., Masud, M. A., Mridha, M. F., Aung, Z., & Dey, N. (2025). KACQ-DCNN: Uncertainty-Aware Interpretable Kolmogorov-Arnold Classical-Quantum Dual-Channel Neural Network for Heart Disease Detection. Computers in Biology and Medicine.
Download Paper