Deep Learning-Based Composite Iterative Reconstruction Algorithm Improves High-Resolution MRI Quality of Hippocampus: A Dual-center Study

Motivation: High-resolution imaging of the hippocampus is essential for diagnosing neurological conditions; however, it is often limited by long scan times and low spatial resolution. Goal(s): To validate the effectiveness of a deep learning-based composite super-resolution reconstruction algorithm combining SRGAN with denoising and sharpening modules to enhance hippocampal image quality without increasing scan time. Approach: 106 patients from two medical centers with hippocampal MRI indications were included. Images were processed using the proposed algorithm. Image quality was assessed using objective and subjective measures. Results: The proposed deep-learning method significantly improved hippocampal structural detail and reduced noise compared to conventional methods. Impact: The deep learning-based composite super-resolution reconstruction method improved 3T MRI hippocampal resolution, allowing better visualization of subtle structures crucial for diagnosing conditions like temporal lobe epilepsy and Alzheimer's disease, without increasing scan time.