Back

Personalized Prediction of Regional Brain Atrophy in Parkinson's Disease through Longitudinal AI Modeling

Wu, Y.; Rundek, T.; Nguyen, P.; Yesha, Y.; Feldman, M.; Atit, P.; Miller, J.; Najafi, T.; Haq, I.

2025-12-30 neurology
10.64898/2025.12.29.25343190 medRxiv
Show abstract

Parkinsons disease (PD) involves variable patterns of brain atrophy in different motor and cognitive regions that differ across patients in both location and progression rate. Existing clinical tools and single-modal imaging methods can be difficult to analyze how individual patients atrophy will progress. Thus, we developed and evaluated deep learning models capable of understanding the pattern of 216 features of PD patients and predicting future regional brain atrophy in individuals with PD using longitudinal 3D MRI and clinical features. Our models combining long short-term memory (LSTM) networks and deep learning classification models to capture time-dependent changes in 29 PD-related brain volumes. Our model demonstrated accurate prediction of regional brain atrophy. In the classification models, 11 regions achieved AUROC greater than 80%. Regression results showed that our model produces an average MAE of 0.4395 and multiple regions R2 are more than 0.8. In conclusion, PD-related regional brain atrophy can be forecast with high test accuracy in longitudinal research cohort. These proof-of-concept results support the feasibility of developing personalized prognostic tools that integrate longitudinal 3D imaging with clinical data.

Matching journals

The top 6 journals account for 50% of the predicted probability mass.

50% of probability mass above

"Similar papers" are the closest papers from that journal in the model's embedding space. They show what the match is built on, but the ranking comes mostly from a classifier over the whole training set, not from these examples alone.