Interpretable LSTM model reveals transiently-realized patterns of dynamic brain connectivity that predict patient deterioration or recovery from very mild cognitive impairment

Yutong Gao, Noah Lewis, Vince D. Calhoun, Robyn L. Miller

Abstract: Highlights•Questionable mild cognitive impairment (qMCI) can be an early-stage of Alzheimer's.•LSTM-based model is trained to predict qMCI recovery vs. deterioration.•An interpretation framework is proposed to understand the model's decision, and,•Explore transient-realized dynamic biomarkers for qMCI prognostic future.•Interpretation framework's trustworthiness is extensively validated in simulations.