Deep Learning for Brain MRI Confirms Patterned Pathological Progression in Alzheimer's Disease

Area of Science:

• Neuroimaging
• Artificial Intelligence
• Neurology

Background:

• Deep learning (DL) excels at differentiating Alzheimer's disease (AD) using brain MRI.
• The utility of DL in detecting progressive structural MRI (sMRI) abnormalities in AD remains underexplored.

Purpose of the Study:

• To propose an interpretable DL algorithm, Ensemble 3DCNN, for investigating longitudinal sMRI changes in AD.
• To identify patterns of neurodegeneration indicative of AD onset and progression using whole-brain sMRI.

Main Methods:

• Utilized 2369 T1-weighted sMRI scans from ADNI and OASIS cohorts.
• Developed and validated an Ensemble of 3D convolutional neural network (Ensemble 3DCNN).
• Generated an Ensemble-3DCNN-based P-score for pattern analysis of neurodegeneration.

Main Results:

• Identified early and connected neurodegeneration in amygdala, insula, parahippocampus, and temporal gyrus.
• Observed complex individual variability in sMRI changes.
• Confirmed AD pathological progression through patterned sMRI changes detected by the DL model.

Conclusions:

• Interpretable DL combined with non-invasive sMRI can detect patterned brain changes associated with AD progression.
• This approach offers new insights into predicting AD progression using whole-brain sMRI data.