Flow-Based Visual Quality Enhancer for Super-Resolution Magnetic Resonance Spectroscopic Imaging

Magnetic Resonance Spectroscopic Imaging (MRSI) is an essential tool for quantifying metabolites in the body, but low spatial resolution limits its clinical applications. Deep learning-based super-resolution methods provided promising results improving of MRSI, super-resolved images are often blurry compared to experimentally-acquired high-resolution images. Attempts have been made with generative adversarial networks improve image visual quality. In this work, we consider another type model, flow-based which training more stable and interpretable networks. Specifically, propose a enhancer network quality MRSI. Different from previous models, our incorporates anatomical information additional modalities (MRI) uses learnable base distribution. addition, impose guide loss data-consistency encourage generate high while maintaining fidelity. Experiments on 1H-MRSI dataset acquired 25 high-grade glioma patients indicate that outperforms baseline methods. Our method also allows adjustment uncertainty estimation. code available at https://github.com/dsy199610/Flow-Enhancer-SR-MRSI .