Optimized Transfer Learning Based Dementia Prediction System for Rehabilitation Therapy Planning.

Area of Science:

• Neuroscience
• Artificial Intelligence
• Medical Imaging

Background:

• Dementia is a progressive neurodegenerative disease impacting cognition and daily function.
• Current treatments can only slow dementia progression, not halt or reverse it.
• Early prediction of dementia is vital for potential prevention and intervention strategies.

Purpose of the Study:

• To develop a novel transfer-learning machine learning model for predicting dementia.
• To utilize magnetic resonance imaging (MRI) data for dementia prediction.
• To improve early diagnosis capabilities for dementia.

Main Methods:

• A transfer-learning machine learning model was developed.
• K-fold cross-validation and parameter optimization were employed for model training.
• Data augmentation using synthetic minority oversampling was utilized.

Main Results:

• The developed model achieved a prediction accuracy of 90.7%.
• The model's performance surpassed competing methods on the same dataset.
• The study demonstrated the efficacy of machine learning in dementia prediction from MRI.

Conclusions:

• The proposed model facilitates early dementia diagnosis, crucial for mitigating neurological decline.
• This AI-driven approach offers a valuable tool for underserved regions lacking physician access.
• Future applications include planning personalized rehabilitation therapy programs for dementia patients.