Empowering Portable Age-Related Macular Degeneration Screening: Evaluation of a Deep Learning Algorithm for a...

Area of Science:

• Ophthalmology
• Artificial Intelligence
• Medical Imaging

Background:

• Large-scale screening for age-related macular degeneration (AMD) remains a global challenge.
• Smartphone-based retinal imaging offers a potential avenue for accessible eye disease detection.

Purpose of the Study:

• To evaluate the performance of an artificial intelligence (AI) algorithm in detecting referable AMD.
• To assess the efficacy of AI in analyzing images from a portable fundus camera.

Main Methods:

• A retrospective dataset from the Age-Related Eye Disease Study (AREDS) was utilized.
• An AI algorithm was trained and fine-tuned on macula-centric retinal images.
• Performance was evaluated using sensitivity, specificity, and Area Under the Curve (AUC).

Main Results:

• The deep learning (DL) algorithm achieved high sensitivity (93.48%) and specificity (82.33%) on the initial AREDS dataset.
• After fine-tuning, the DL algorithm demonstrated a sensitivity of 91.25% and specificity of 84.18% on images from the target device.
• The AUC remained high, indicating strong diagnostic performance for referable AMD detection.

Conclusions:

• The DL algorithm shows significant promise for detecting referable AMD using smartphone-based imaging.
• This AI-driven approach could facilitate cost-effective and widespread AMD screening, particularly in underserved regions.
• Further validation on diverse populations is warranted to confirm its clinical utility.