Automated detection and classification of early AMD biomarkers using deep learning

Area of Science:

• Ophthalmology
• Medical Imaging
• Artificial Intelligence

Background:

• Age-related macular degeneration (AMD) is a primary cause of vision loss globally.
• Early detection of AMD is crucial to prevent irreversible vision impairment.
• Current OCT-based biomarker staging is time-consuming and requires specialized expertise.

Purpose of the Study:

• To develop a deep learning method for automated detection and classification of early AMD biomarkers from OCT scans.
• To improve the efficiency and accessibility of AMD staging in screening settings.

Main Methods:

• Utilized deep convolutional neural networks (CNNs) trained on approximately 20,000 OCT B-scans from 153 patients.
• Focused on automated detection of hyperreflective foci, hyporeflective foci within drusen, and subretinal drusenoid deposits.
• Evaluated various state-of-the-art CNNs and transfer learning protocols.

Main Results:

• Achieved an overall accuracy of 87% in identifying early AMD biomarkers.
• Demonstrated the potential of CNNs for accurate AMD biomarker classification.
• The method shows promise for integration into screening workflows.

Conclusions:

• Automated deep learning analysis of OCT scans can effectively detect early AMD biomarkers.
• This approach can potentially streamline AMD staging and facilitate earlier patient management.
• Further development could enhance diagnostic capabilities for ophthalmologists and researchers.