Quantum Neural Network for Accelerated Magnetic Resonance Imaging
Shuo Zhou, Yihang Zhou, Congcong Liu, Yanjie Zhu, Hairong Zheng, Dong Liang, Haifeng Wang
TL;DR
MRI reconstruction from undersampled data is challenging due to nonlinear feature recovery. The authors propose a hybrid quantum-classical network that maps zero-filled MR images $z$ to fully sampled images $y$ using a quantum convolution front-end and a classical U‑net back-end, trained end-to-end under the objective $ \min_{\Theta} \sum_{t=1}^T || C_{unet}( U(z_t); \Theta ) - y_t ||_2^2$. Contributions include a 4-qubit, 2×2 quantum convolution with angle encoding and CNOT entanglement, integration with an asymmetric U‑net, and an end-to-end training framework showing comparable performance at $2\times$ acceleration and superior performance at $4\times$. The study demonstrates the feasibility and potential advantages of quantum-enhanced MR reconstruction and points toward hardware validation and extension to in vivo imaging in the future.
Abstract
Magnetic resonance image reconstruction starting from undersampled k-space data requires the recovery of many potential nonlinear features, which is very difficult for algorithms to recover these features. In recent years, the development of quantum computing has discovered that quantum convolution can improve network accuracy, possibly due to potential quantum advantages. This article proposes a hybrid neural network containing quantum and classical networks for fast magnetic resonance imaging, and conducts experiments on a quantum computer simulation system. The experimental results indicate that the hybrid network has achieved excellent reconstruction results, and also confirm the feasibility of applying hybrid quantum-classical neural networks into the image reconstruction of rapid magnetic resonance imaging.