CQ-CNN: A lightweight hybrid classical-quantum convolutional neural network for Alzheimer's disease detection using...

Area of Science:

• Medical Imaging
• Quantum Computing
• Artificial Intelligence

Background:

• Alzheimer's disease (AD) detection from 3D MRI data is complex, traditionally using classical CNNs.
• The rise of quantum computing necessitates exploring quantum-based systems for AD detection.
• Current classical-quantum architectures require investigation for their capabilities and limitations.

Purpose of the Study:

• To develop automated Alzheimer's disease detection systems utilizing quantum computing.
• To evaluate the performance and limitations of hybrid classical-quantum neural networks for AD detection.
• To introduce a novel preprocessing framework and a quantum-enhanced CNN model.

Main Methods:

• A preprocessing framework converting 3D MRI data to 2D slices.
• CQ-CNN: a lightweight hybrid classical-quantum convolutional neural network using parameterized quantum circuits (PQC).
• Experiments conducted on the OASIS-2 dataset.

Main Results:

• Hybrid classical-quantum models face convergence issues with similar image classes (e.g., moderate dementia vs. non-dementia), leading to gradient failure.
• When convergence is achieved, the quantum model shows a quantum advantage, reaching state-of-the-art accuracy with significantly fewer parameters.
• A 7-3-qubit model achieved 97.5% accuracy with 13.7K parameters, outperforming a classical model by 5.67%.

Conclusions:

• Current hybrid classical-quantum architectures have limitations in optimization for AD detection.
• Quantum models offer a promising quantum advantage in accuracy and parameter efficiency for medical imaging.
• Further advancements in quantum optimization methods are crucial for practical deployment in AD detection.