A new weakly supervised deep neural network for recognizing Alzheimer's disease.

Area of Science:

• Neuroscience
• Medical Imaging
• Artificial Intelligence

Background:

• Alzheimer's disease (AD) is a neurodegenerative condition impacting cognition and memory, primarily in older adults.
• Machine learning aids AD diagnosis, but supervised methods struggle with low-quality or unlabeled medical data.
• High costs associated with labeling large medical datasets limit the application of traditional supervised learning.

Purpose of the Study:

• To propose a Weakly Supervised Deep Learning (WSDL) model for Alzheimer's disease diagnosis.
• To address challenges posed by unlabeled and low-quality medical data in AD prediction.
• To enhance the utilization of extensive unlabeled medical datasets for improved diagnostic accuracy.

Main Methods:

• Developed a WSDL model integrating attention mechanisms and consistency regularization within the EfficientNet framework.
• Employed data augmentation techniques to maximize the utility of unlabeled data.
• Validated the WSDL model on Alzheimer's Disease Neuroimaging Program (ADNI) brain MRI datasets with varying unlabeled data ratios.

Main Results:

• The proposed WSDL method demonstrated superior performance compared to baseline methods in Alzheimer's disease diagnosis.
• Experimental results confirmed the model's effectiveness across different ratios of unlabeled data.
• The WSDL approach successfully utilized unlabeled brain MRI data for robust AD prediction.

Conclusions:

• The WSDL model offers a promising solution for Alzheimer's disease diagnosis, particularly when dealing with limited labeled data.
• This approach effectively harnesses unlabeled medical imaging data, reducing reliance on costly manual labeling.
• The findings highlight the potential of weakly supervised learning in advancing neurodegenerative disease diagnostics.