FundusNet: A Deep-Learning Approach for Fast Diagnosis of Neurodegenerative and Eye Diseases Using Fundus Images

Area of Science:

• Ophthalmology
• Neurology
• Artificial Intelligence

Background:

• Early detection of neurodegenerative diseases is critical for treatment.
• Current diagnostic methods are often expensive and time-consuming.

Purpose of the Study:

• To introduce FundusNet, a deep learning model for rapid and cost-effective diagnosis of neurodegenerative diseases using fundus images.
• To evaluate FundusNet's performance in age prediction, gender classification, and diagnosing specific neurodegenerative diseases.

Main Methods:

• A deep learning model, FundusNet, was trained on fundus images.
• Performance was assessed using metrics like Mean Absolute Error (MAE) and Area Under the Curve (AUC).
• Grad-CAM was employed for model interpretability, identifying image regions crucial for diagnosis.

Main Results:

• FundusNet achieved high accuracy in age prediction (MAE 2.55 years) and gender classification (AUC 0.98).
• The model demonstrated significant performance in diagnosing Parkinson's disease (AUC 0.75 ± 0.03) and multiple sclerosis (AUC 0.77 ± 0.02).
• Grad-CAM analysis confirmed FundusNet's focus on relevant retinal structures, and genetic risk prediction showed high accuracy.

Conclusions:

• FundusNet provides a rapid, cost-effective tool for early neurodegenerative disease detection via fundus imaging.
• The model's interpretability highlights its ability to identify disease-associated retinal biomarkers.
• Further enhancement is possible with larger datasets, particularly for predicting genetic risk.