Forecasting the progression of Alzheimer's disease using neural networks and a novel preprocessing algorithm

Area of Science:

• Neurology
• Artificial Intelligence
• Biomedical Informatics

Background:

• Alzheimer's disease (AD) clinical trials face a 99.6% failure rate, largely due to difficulties in early patient identification.
• Early and accurate diagnosis is crucial for effective intervention and therapeutic development in AD.

Purpose of the Study:

• To investigate machine learning (ML) methods for predicting AD progression using patient clinical data.
• To develop a predictive model capable of identifying individuals at early stages of Alzheimer's disease.

Main Methods:

• Utilized the novel "All-Pairs" technique to process temporal data from 1737 patients.
• Trained and evaluated machine learning models, including a neural network, on processed patient data.
• Validated model performance using a separate testing dataset of 110 patients.

Main Results:

• A neural network model achieved significant predictive accuracy (mAUC = 0.866) for AD progression.
• The model demonstrated effectiveness in predicting progression for both cognitively normal individuals and those with mild cognitive impairment.

Conclusions:

• The developed ML model can identify patients in the early stages of AD.
• This predictive capability can enhance the selection of suitable candidates for Alzheimer's disease therapeutic clinical trials.