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

Published in ICASSP 2025 – IEEE International Conference on Acoustics, Speech, and Signal Processing (ICASSP), 2025

CompressedMediQ presents a scalable hybrid quantum-classical pipeline tailored to the demands of high-dimensional neuroimaging data, particularly for dementia staging. The pipeline features:

  • Classical HPC-powered pre-processing and CNN-PCA-based feature extraction to address data dimensionality constraints.
  • Quantum Support Vector Machine (QSVM) classification for optimized feature mapping and superior separability.
  • Application to real-world datasets (e.g., ADNI and NIFD), demonstrating higher accuracy than classical-only methods.

Despite limitations in qubit availability under NISQ constraints, CompressedMediQ validates the potential of quantum-enhanced diagnostics in healthcare, advancing the future of signal processing and neuroimaging analysis.

Figure: CompressedMediQ pipeline integrating classical pre-processing, QSVM classification, and neuroimaging data from hospital MRI.

Recommended citation: Chen, Kuan-Cheng; Li, Yi-Tien; Li, Tai-Yu; Liu, Chen-Yu; Lee, Po-Heng; & Chen, Cheng-Yu. (2025). "CompressedMediQ: Hybrid Quantum Machine Learning Pipeline for High-Dimensional Neuroimaging Data." Proceedings of ICASSP 2025 – IEEE International Conference on Acoustics, Speech, and Signal Processing.
Download Paper