What question did this study set out to answer?

To enhance clinical diagnostics via adaptive quantum kernel methods that optimize data handling and feature mapping, addressing limitations in classical approaches.

June 14, 2026Open Access

Adaptive quantum kernel selection via leakage-free stacking for clinical diagnostics on NISQ hardware

Key Points

To enhance clinical diagnostics via adaptive quantum kernel methods that optimize data handling and feature mapping, addressing limitations in classical approaches.
Utilized three distinct quantum feature maps (Angle, Amplitude, ZZ-entanglement) for mapping clinical data.
Implemented a meta-learner to optimize kernel selection through nested cross-validation (5-fold inner, 10-fold outer).
Conducted evaluations on datasets for Parkinson’s, Breast Cancer, and Diabetes with 1,000-iteration bootstrapping.
Raised Parkinson’s Specificity from 0.585 to 0.813 (p<0.001) while maintaining Recall above 0.95.
Matched classical RBF-SVM on Breast Cancer (all p>0.05).
Improved Diabetes Recall from 0.553 to 0.621 (p=0.004).

Abstract

Quantum kernel methods map clinical features into exponentially large Hilbert spaces where overlapping biological markers can become more separable than in fixed-dimensional classical feature spaces, but existing work evaluates single quantum feature maps, ignores barren-plateau failure modes, and relies on single train–test splits vulnerable to data leakage. Classical diagnostics for Parkinson’s disease, breast cancer, and diabetes remain limited by the Specificity–Recall trade-off that fixed-dimensional kernels impose on overlapping biomarker distributions. We propose an adaptive hybrid quantum framework routing clinical data through three distinct quantum feature maps, namely Angle, Amplitude, and ZZ-entanglement, computing fidelity-based Gram matrices for Quantum SVM and Quantum KNN classifiers. A Logistic Regression meta-learner, trained on strictly out-of-fold predictions from nested cross-validation (5-fold inner, 10-fold outer), learns which quantum kernel generalizes on each dataset and suppresses those that do not. Evaluated on Parkinson’s (195 patients), Breast Cancer (569), and Diabetes (768) with 1,000-iteration bootstrapping, the ensemble raised Parkinson’s Specificity from 0.585 to 0.813 ( \(p0.05\) ), and improved Diabetes Recall from 0.553 to 0.621 ( \(p=0.004\) ). A standalone Variational Quantum Classifier failed on all datasets (ROC-AUC 0.51 to 0.57), confirming barren plateau limitations. Explainability via SHAP, LIME, and Permutation Importance revealed dataset-dependent kernel trust: the meta-learner suppressed QKNN Amplitude on Diabetes (coefficient \(-0.353\) ) while amplifying it on Parkinson’s (1.761). PCA-based feature backtracking recovered established biomarkers including Insulin and Glucose for Diabetes, vocal perturbation measures for Parkinson’s, and nucleus geometry for Breast Cancer. Noise simulations confirmed graceful degradation under NISQ conditions. The framework performs data-driven kernel selection, removing the need to pre-specify an encoding strategy.

Ask AI

Mark Helpful

Bookmark

Relay

View Full Paper