Development of Hybrid Classical–quantum Framework for Lossy Compression of 2D Magnetic Resonance Images using a Quantum Autoencoder (QAE)

Abstract

Recent advancements in both MRI scanners’ hardware methods have pushed further the capabilities of medical imaging but also opened new challenges in terms of storage and sharing requirements of ever larger MRI datasets .This process creates large datasets crucial for diagnosis, posing storage and transmission challenges in digital health systems, especially for molecular imaging sequences like diffusion‐weighted MRI (DW-MRI) and dynamic contrast‐enhanced (DCE) MRI. Traditional compression methods often lose important clinical details, highlighting the need for advanced techniques [2–4] that preserve diagnostic accuracy while achieving high compression ratios. Classical compression techniques risk losing subtle diffusion or perfusion signatures crucial for diagnostics. Quantum autoencoders (QAEs) promise efficient representation of high-dimensional data by embedding classical preprocessed features into a small quantum latent space. In this work, we integrate principal component analysis (PCA) with a QAE to compress single 2D MRI slices, enabling potential acceleration in molecular imaging pipelines.