Deep learning maps brain surface and quality from diverse mri scans
Uncertainty-Aware Cortical Surface Reconstruction and Spherical Registration for Heterogeneous Brain MRI
Computer Vision and Pattern Recognition
Summary
It’s hard to map the brain’s surface from MRI scans when the scans vary in quality and type. The authors present a deep learning method trained on synthetic scans that can work with almost any MRI, including those with lower resolution or different imaging contrasts. Their method also tells how certain the brain surface predictions are, helping identify scans that might be less reliable. This approach works as well as existing methods on standard scans and better on challenging ones, and it quickly fits brain surfaces into a common framework for comparison.
What this means in practice
- •For clinical imaging teams: Process diverse clinical MRI scans to extract precise brain surface models despite varying quality and scan types.
- •For medical software developers: Build brain analysis tools that integrate fast surface mapping and quality monitoring across multiple MRI formats.$Commercial implications: Enables creation of adaptable brain analysis software sold to hospitals for heterogeneous MRI data interpretation.
Authors
Jesper Duemose Nielsen, Karthik Gopinath, Andrew Hoopes, Adrian Dalca, Axel Thielscher, Juan Eugenio Iglesias, Oula Puonti
Abstract
Surface-based cortical analysis is useful in many applications as it is a natural way to model the highly folded structure of the cerebral cortex, allowing quantities such as curvature and thickness to be easily calculated. Existing deep learning (DL) methods have focused on 1x1x1 mm3 T1-weighted (T1w) scans, thus replicating classical pipelines. Here we present a neural network that is trained exclusively on synthetic scans with random contrast and resolution. Consequently, it can be applied to almost any kind of magnetic resonance imaging (MRI) scan. In addition, our model provides vertex-wise uncertainty estimates of the predicted surface as well as a spherical registration to fsaverage space. Our model performs on par with other state-of-the-art DL methods on 1x1x1 mm3 T1w scans and outperforms the only other method capable of estimating cortical surfaces from low resolution, non-T1w scans. The predicted uncertainty, moreover, correlates with overall scan quality on low resolution scans and highlights areas of poor surface reconstruction on 1x1x1 mm3 T1w scans, suggesting that it can be used for quality control. Finally, the predicted spherical registrations prove comparable to other DL and classical registration methods. These qualities, when combined with the fast inference time, will provide the ability to analyze large, existing databases of clinical MRI scans, with the potential for finding disease-specific patterns that would otherwise be infeasible in a research setting.