Automated Alzheimer's disease detection using active learning model with reinforcement learning and scope loss...

Area of Science:

• Neurology
• Artificial Intelligence
• Medical Imaging

Background:

• Alzheimer's disease (AD) detection is crucial for management but traditional methods require extensive labeled image data.
• Current active learning methods for AD detection often lack adaptability due to static data selection strategies.

Purpose of the Study:

• To develop an innovative active learning framework for early Alzheimer's disease detection that requires fewer labeled samples.
• To enhance model performance and adaptability in AD detection using advanced machine learning techniques.

Main Methods:

• A novel active learning framework combining deep reinforcement learning (DRL) with a scope loss function (SLF) for dynamic data selection.
• Integration of a differential evolution (DE) algorithm to mitigate hyperparameter sensitivity in the DRL component.
• Evaluation on the OASIS and ADNI datasets for Alzheimer's disease detection.

Main Results:

• The proposed framework demonstrated superior performance in early Alzheimer's disease detection compared to conventional methods.
• Achieved high F-measures of 92.044% on the OASIS dataset and 93.685% on the ADNI dataset.
• The dynamic data selection strategy effectively balanced data exploitation and exploration, improving model efficiency.

Conclusions:

• The developed active learning framework offers a more efficient and adaptable approach to early Alzheimer's disease detection.
• This method significantly reduces the need for large, pre-labeled datasets, addressing a key challenge in medical image analysis.
• The combination of DRL, SLF, and DE presents a promising advancement for neurodegenerative disease diagnostics.