Open-source and data-driven hippocampal research

HippUnfold and HippoMaps provide automated image analysis and informatics, respectively. These tools are the backbone of the lab, and have spawned many major recent directions including:

• multimodal imaging spanning MRI at the millimeter scale to histology microscale
• cross-species translation (human-macaque-marmoset-rodent)
• crossing gene mapping with neuroimaging
• simulations using neural mass and neural field models

AI: what are we doing with it?

• Our image processing pipelines employ AI to squeeze out the most information from the data, whether it be standard clinical MRI scans, specialized high-resolution scans, or even textural feature extraction from histological images. As such, we are always looking for new ways to improve our pipelines, and are open to collaborations with AI experts.
• We are also interested in translation between hippocampal cognition at the microcircuit level and AI architectures. In particular, recent cognitive models of the hippocampus emphasize its involvement in prediction processes which can provide critical training signals or cost functions for AI models. This is an exciting direction in the AI community since it helps shift the rate-limiter away from having abundant training data and more towards naturalistic exploration of an unlabelled virtual or real world. Read more about it in our upcoming Perspective article (coming soon!).

What's special about our analytic approach?

• We leverage the notion that the hippocampus is cortical - that is, its composed of a thin and highly folded archicortical sheet. This allows us to use surface-based analysis techniques to map the hippocampus in 3D, and to compare it across individuals, scales, species, and modalities. This is a powerful approach because it allows us to compare the hippocampus in a way that is not biased by the size of the structure, and is more sensitive to subtle changes in shape or volume. This is particularly important in the context of subtle differences such as those seen in neurodivergence, psychiatric disorders, and early-stage neurodegenerative diseases, where the hippocampus is often one of the first structures to be affected.
• This also allows for mapping of data onto standardized surfaces (or unfolded space) with vertex-wise equivalence across individuals. This provides a way to share data at an intermediate level of analysis, and to compare data across studies. This is particularly important in the context of large-scale studies, where data sharing is critical for increasing statistical power.

Where are we?

Jordan DeKraker is housed as a postdoc within the MCIN and MICA labs of Dr. Alan Evans and Dr. Boris Bernhardt, respectively, at the Montreal Neurological Institute, McGill University. The labs are part of the Healthy Brains for Healthy Lives initiative, and are affiliated with the McConnell Brain Imaging Centre, the Ludmer Centre for Neuroinformatics and Mental Health, and the Douglas Mental Health University Institute. The labs are also part of the Canadian Open Neuroscience Platform, and are involved in the development of the Canadian Brain Imaging Research Platform.

Jordan is looking for a permanent position to establish this lab further! Please checkout the Contacts/Links page if you are looking to hire, join, or to collaborate.