PRETTIER-MRI
Code to perform super-resolution on structural MRI using fine-tuned convolutional neural networks.
Description:
Perceptual super-REsoluTion in mulTIple sclERosis (PRETTIER) MRI is an approach to perform single-image super-resolution on structural MRI, specifically T2-W FLAIR and T1-W MRIs of people with Multiple Sclerosis. This framework relies on convolutional neural networks that have been fine-tuned with a content loss function that promotes perceptual quality and reconstruction accuracy to recover high-level image features. Evaluation of PRETTIER MRI shows that it leads to more accurate and faster MRI reconstructions than competing methods. Furthermore, it improves the automated segmentation of Multiple sclerosis lesions on low-resolution MRI, approaching the performance achievable with high-resolution images. PRETTIER MRI addresses the limitations of multi-slice MRI acquired in real-world clinical settings, facilitating the use of real-world MRI datasets to investigate quantitative image-based biomarkers of disease outcomes.
More Information:
Originates from the AI-Driven Data Science research challenge (FAIR 2019–2023).
Responsible Research Lead: Jan Sijbers
Authors: Diana L. Giraldo
Programming Language: Python
Categories: Images, Supervised learning