CompressedMediQ: Hybrid Quantum Machine Learning Pipeline for High-Dimensional Neuroimaging Data

TL;DR

This work tackles the challenge of classifying dementia stages from high-dimensional MRI data by introducing CompressedMediQ, a hybrid quantum-classical pipeline that couples HPC-based MRI pre-processing and CNN-PCA feature extraction with QSVM-powered quantum kernels. The approach compresses features to a tractable size (8 dimensions) before quantum encoding, enabling efficient quantum classification in the NISQ era. Experimental results on ADNI and NIFD data demonstrate superior accuracy and reduced misclassification for the quantum model, particularly in early dementia stages, highlighting the practical potential of quantum-enhanced learning in clinical diagnostics. The study lays groundwork for scalable quantum-assisted neuroimaging tools, while outlining future work on error mitigation, distributed quantum computing, and resource management to handle larger datasets.

Abstract

This paper introduces CompressedMediQ, a novel hybrid quantum-classical machine learning pipeline specifically developed to address the computational challenges associated with high-dimensional multi-class neuroimaging data analysis. Standard neuroimaging datasets, such as large-scale MRI data from the Alzheimer's Disease Neuroimaging Initiative (ADNI) and Neuroimaging in Frontotemporal Dementia (NIFD), present significant hurdles due to their vast size and complexity. CompressedMediQ integrates classical high-performance computing (HPC) nodes for advanced MRI pre-processing and Convolutional Neural Network (CNN)-PCA-based feature extraction and reduction, addressing the limited-qubit availability for quantum data encoding in the NISQ (Noisy Intermediate-Scale Quantum) era. This is followed by Quantum Support Vector Machine (QSVM) classification. By utilizing quantum kernel methods, the pipeline optimizes feature mapping and classification, enhancing data separability and outperforming traditional neuroimaging analysis techniques. Experimental results highlight the pipeline's superior accuracy in dementia staging, validating the practical use of quantum machine learning in clinical diagnostics. Despite the limitations of NISQ devices, this proof-of-concept demonstrates the transformative potential of quantum-enhanced learning, paving the way for scalable and precise diagnostic tools in healthcare and signal processing.

Figures (5)

• Figure 1: Overview of the CompressedMediQ Pipeline for High-Dimensional Neuroimaging Data Analysis. The pipeline integrates classical and quantum computing nodes to process MRI data for disease classification. The workflow begins with data pre-processing on HPC classical nodes, including segmentation and voxel-based morphometry analysis, followed by feature extraction using CNNs and dimensionality reduction via PCA. The extracted features are then input into the quantum machine learning stage, where quantum kernel estimation and a QSVM are employed to perform multi-class classification of disease stages.
• Figure 2: Overview of the MRI Preprocessing Procedure and VBM Analysis, illustrating the steps from image correction and segmentation to VBM map generation, and nonlinear warping of segmented tissue probability maps to standardized template space for preserving volumetric information of GM, WM, and CSF.
• Figure 3: Overview of the CompressedMediQ pipeline illustrating the hybrid quantum-classical workflow, including data preprocessing with CNN and PCA, quantum kernel estimation, and kernel learning using QSVM for enhanced multi-class classification of high-dimensional MRI data.
• Figure 4: The hybrid framework integrates MRI-derived features with demographic, genomic, and clinical history inputs into a combined regressor/classifier model for enhanced neuroimaging analysis and prediction.
• Figure 5: Confusion matrices comparing Quantum (a) and Classical SVM (b) models, highlighting superior accuracy and reduced misclassification in the Quantum model (96.1%) versus the Classical SVM model (78.8%), particularly within the marked misclassification region.