Comparative Analysis of Automated vs. Expert-Designed Machine Learning Models in Age-Related Macular Degeneration...

Area of Science:

• Ophthalmology
• Artificial Intelligence
• Medical Imaging

Background:

• Age-related macular degeneration (AMD) is a leading cause of vision loss.
• Accurate classification of AMD from optical coherence tomography (OCT) images is crucial for timely treatment.
• Machine learning offers potential for automated image analysis.

Purpose of the Study:

• To compare the diagnostic performance of expert-designed machine learning models versus code-free automated machine learning (AutoML) models.
• To evaluate the effectiveness in classifying normal OCT images, detecting AMD, and distinguishing between dry and wet AMD forms.

Main Methods:

• Expert-designed models utilized EfficientNet V2 architecture.
• AutoML models were developed using LobeAI with ResNet-50 V2 transfer learning.
• Both models were trained and tested on 500 OCT images per diagnostic group with an 80:20 split.

Main Results:

• Expert-designed models achieved 99.67% overall accuracy and F1 scores ≥0.99.
• AutoML models achieved 89.00% overall accuracy and F1 scores ranging from 0.86 to 0.90.
• AutoML models showed lower recall for identifying dry AMD.

Conclusions:

• Expert-designed models significantly outperformed AutoML models in AMD classification.
• The study emphasizes the necessity of expert involvement for high-precision medical image analysis.
• Advanced architectures and optimization by AI experts yield superior diagnostic tools.