Advances in MRI Harmonization and Analysis

The field of neuroimaging is moving towards developing more robust and generalizable solutions for large-scale multicenter studies. Researchers are focusing on creating innovative frameworks for MRI harmonization, which can reduce inter-site variability and improve downstream model performance. Another area of interest is the development of foundation models for 3D brain MRI, which can enable general-purpose feature learning from large-scale, unlabeled datasets. These models have shown promise in improving diagnostic accuracy and reducing dependency on extensive expert annotations. Noteworthy papers in this area include: Scanner-Agnostic MRI Harmonization via SSIM-Guided Disentanglement, which presents a novel image-based harmonization framework for 3D T1-weighted brain MRI. Towards Generalisable Foundation Models for 3D Brain MRI, which introduces BrainFound, a self-supervised foundation model for brain MRI that consistently outperforms existing self-supervised pretraining strategies and supervised baselines.

Sources

Scanner-Agnostic MRI Harmonization via SSIM-Guided Disentanglement

Towards Generalisable Foundation Models for 3D Brain MRI

Machine Learning based Analysis for Radiomics Features Robustness in Real-World Deployment Scenarios

Auto3DSeg for Brain Tumor Segmentation from 3D MRI in BraTS 2023 Challenge

Fine-tuning Segment Anything for Real-Time Tumor Tracking in Cine-MRI

Built with on top of