Reconstructing and forecasting disease trajectories of patients with Alzheimer's disease using routine data in resource-constrained settings
Researchers have developed a new framework called GNOVA, which uses a GRU-Neural ODE Variational Autoencoder to predict and reconstruct Alzheimer's disease progression. This model can forecast cognitive scores like CDR-SB and MMSE using only routine patient data, avoiding the need for expensive neuroimaging or biomarker tests. GNOVA achieved low mean absolute errors on a dataset of 1,727 patients, demonstrating its potential for clinical decision-making in resource-constrained environments. AI
IMPACT Enables more accessible and cost-effective patient monitoring and prognosis for Alzheimer's disease.