A Comparative Multi-Modality Evaluation of Ensemble Machine Learning and Variational Quantum Classification for Alzheimer’s Disease Prediction

Abstract

Alzheimer’s disease prediction requires robust mod eling of heterogeneous medical data, including structural MRI representations and structured clinical attributes. This study presents a controlled multi-modality benchmarking framework comparing optimized classical ensemble learning methods with a Variational Quantum Classifier (VQC) under identical prepro cessing and validation protocols. MRI features are reduced using Principal Component Analysis (PCA), while structured clinical attributes are modeled using Random Forest, XGBoost, Voting, and Stacking ensembles. A hybrid quantum–classical pipeline is implemented using Qiskit and PennyLane to evaluate near-term quantum feasibility under NISQ constraints. Experimental results demonstrate that stacking ensemble mod els achieve 95.3% accuracy on clinical data and 91.5% on MRI data, significantly outperforming the VQC, which achieves 71.4% accuracy under the same evaluation conditions. Statistical testing confirms that this performance gap is significant. These findings indicate that optimized classical ensemble learn ing remains superior for current medical prediction tasks, while variational quantum classification remains exploratory under present hardware limitations.

Keywords: Alzheimer’s Disease, Multi-Modal Learning, Ensemble Learning, Variational Quantum Classification, Hybrid Quantum-Classical Learning, Medical Image Analysis, Clinical Data Analytics, Noisy Intermediate-Scale Quantum (NISQ) Com puting.