Exploring the potential of representation and transfer learning for anatomical neuroimaging: Application to psychiatry.

The perspective of personalized medicine for brain disorders requires efficient learning models for anatomical neuroimaging-based prediction of clinical conditions. There is now a consensus on the benefit of deep learning (DL) in addressing many medical imaging tasks, such as image segmentation. However, for single-subject prediction problems, recent studies yielded contradictory results when comparing DL with Standard Machine Learning (SML) on top of classical feature extraction. Most existing comparative studies were limited in predicting phenotypes of little clinical interest, such as sex and age, and using a single dataset. Moreover, they conducted a limited analysis of the employed image pre-processing and feature selection strategies. This paper extensively compares DL and SML prediction capacity on five multi-site problems, including three increasingly complex clinical applications in psychiatry namely schizophrenia, bipolar disorder, and Autism Spectrum Disorder (ASD) diagnosis. To compensate for the relative scarcity of neuroimaging data on these clinical datasets, we also evaluate three pre-training strategies for transfer learning from brain imaging of the general healthy population: self-supervised learning, generative modeling and supervised learning with age. Overall, we find similar performance between randomly initialized DL and SML for the three clinical tasks and a similar scaling trend for sex prediction. This was replicated on an external dataset. We also show highly correlated discriminative brain regions between DL and linear ML models in all problems. Nonetheless, we demonstrate that self-supervised pre-training on large-scale healthy population imaging datasets (N≈10k), along with Deep Ensemble, allows DL to learn robust and transferable representations to smaller-scale clinical datasets (N≤1k). It largely outperforms SML on 2 out of 3 clinical tasks both in internal and external test sets. These findings suggest that the improvement of DL over SML in anatomical neuroimaging mainly comes from its capacity to learn meaningful and useful abstract representations of the brain anatomy, and it sheds light on the potential of transfer learning for personalized medicine in psychiatry.

OpenBHB: a Large-Scale Multi-Site Brain MRI Data-set for Age Prediction and Debiasing.

Prediction of chronological age from neuroimaging in the healthy population is an important issue because the deviations from normal brain age may highlight abnormal trajectories towards brain disorders. As a first step, ML models have emerged to predict chronological age from brain MRI, as a proxy measure of biological age. However, there is currently no consensus w.r.t which Machine Learning (ML) model is best suited for this task, largely because of a lack of public benchmark. Furthermore, new large emerging population neuroimaging datasets are often biased by the acquisition center images are coming from. This bias heavily deteriorates models generalization capacities, especially for Deep Learning (DL) algorithms that are known to overfit rapidly on the simplest features (known as simplicity bias). Here we propose a new public benchmarking resource, namely Open Big Healthy Brains (OpenBHB), along with a challenge for both brain age prediction and site-effect removal through a representation learning framework. OpenBHB is large-scale, gathering >5K 3D T1 brain MRI from Healthy Controls (HC) and highly multi-sites, aggregating >60 centers worldwide and 10 studies. OpenBHB is expected to grow both in terms of available modalities and number of subjects. All OpenBHB datasets are uniformly preprocessed, including quality check, with container technologies that consist in: 3D Voxel-Based Morphometry maps (VBM from CAT12), quasi-raw (simple linear alignment of images), and Surface-Based Morphometry indices (SBM, from FreeSurfer). The OpenBHB challenge is permanent and we provide all tools, materials and tutorials for participants to easily submit and benchmark their model against each other on a public leaderboard.