A modified deep learning method for Alzheimer's disease detection based on the facial submicroscopic features in mice

Area of Science:

• Biomedical Engineering
• Artificial Intelligence in Medicine
• Neuroscience

Background:

• Alzheimer's disease (AD) is a growing global public health concern, particularly with aging populations.
• Early diagnosis of AD is crucial for effective treatment strategies.
• Identifying reliable early detection methods is a significant challenge in AD research.

Purpose of the Study:

• To develop and refine a deep learning model for identifying submicroscopic, AD-related characteristics in mouse facial images.
• To evaluate the model's performance in distinguishing between normal and AD-affected mice.
• To explore the potential of facial imaging as a non-invasive method for AD detection.

Main Methods:

• Development of a multi-layer cyclic Residual convolutional neural network (CNN).
• Classification of mice into normal control and AD groups based on facial image analysis.
• Utilizing Class Activation Mapping (CAM) to visualize AD-related features in facial images.

Main Results:

• The proposed CNN model demonstrated superior detection performance compared to other deep learning models.
• Achieved high accuracy (99.78%), sensitivity (100%), specificity (99.65%), and precision (99.44%) in AD identification.
• CAM analysis confirmed the presence of AD-related submicroscopic features in mouse facial images.

Conclusions:

• Facial image-based deep learning models are feasible and accurate for identifying Alzheimer's disease in mice.
• The study validates the potential of using facial images for early AD detection in humans via deep learning.
• This research opens new avenues for non-invasive diagnostic tools in Alzheimer's disease research.