Age Sensitive Hippocampal Functional Connectivity: New Insights from 3D CNNs and Saliency Mapping.

Area of Science:

• Neuroscience
• Artificial Intelligence
• Medical Imaging

Background:

• Hippocampal grey matter loss is a key indicator of neurobiological aging.
• Functional connectivity (FC) changes in the aging hippocampus are not fully understood.
• Seed-based FC analysis maps synchronous activity between the hippocampus and cortical regions.

Purpose of the Study:

• To develop an interpretable deep learning framework for predicting brain age from hippocampal FC.
• To identify specific hippocampal-cortical connections sensitive to aging.
• To investigate age-related functional reorganization in the hippocampus.

Main Methods:

• Utilized a three-dimensional convolutional neural network (3D CNN) for brain age prediction.
• Employed LayerCAM saliency mapping for interpretable analysis of hippocampal FC.
• Disaggregated anterior and posterior hippocampal FC to explore functional specializations.

Main Results:

• Identified significant age-related hippocampal-cortical connections, notably with the precuneus, cuneus, and posterior cingulate cortex.
• Highlighted connections with the parahippocampal cortex, left superior parietal lobule, and right superior temporal sulcus as age-sensitive.
• Demonstrated distinct FC mapping for anterior and posterior hippocampus, correlating with known functions.

Conclusions:

• The study provides novel insights into the functional mechanisms underlying hippocampal aging.
• Explainable deep learning effectively uncovers biologically meaningful patterns in neuroimaging data.
• The developed framework offers a powerful tool for studying brain aging and related disorders.