Deep Image Prior-Based Restoration with Spatial Resolution Enhancement and Noise Reduction in Mouse Cardiac Magnetic Resonance Imaging

초록

Cardiac magnetic resonance imaging (MRI) in small animals is a powerful, non-invasive modality for assessing cardiac structure and function. However, acquiring high-quality images is challenging due to the small heart size and rapid heart rate in mouse models. In this study, we propose a deep image prior (DIP)-based image enhancement algorithm and evaluate its feasibility for four-dimensional (4D) mouse cardiac MR images. The proposed algorithm was modeled using a hybrid stopping strategy. Compared with the conventional approach combining block-matching and 4D filtering with total generalized variation, the proposed approach demonstrated improvements of 31.63%, 146.77%, and 63.16% in contrast-to-noise ratio, gradient magnitude, and perceptual sharpness index, respectively. In conclusion, the proposed DIP-based restoration framework with a hybrid stopping strategy shows potential for improving image quality in small-animal cardiac MRI by jointly reducing noise, mitigating blur, and enhancing apparent spatial resolution under an explicitly defined degradation model.

키워드