DeepLensNet

DeepLensNet automates quantitative detection and classification of age-related cataracts from anterior segment and retroillumination photographs for ophthalmic assessment and research.

Key Features:

• Automated Detection and Quantification: Identifies and measures Nuclear Sclerosis (NS) on a scale of 0.9–7.1 from 45-degree slit-lamp photographs, and quantifies Cortical Lens Opacity (CLO) and Posterior Subcapsular Cataract (PSC) as percentages (0%–100%) from retroillumination photographs.
• Performance Metrics: Evaluated using mean squared error (MSE), achieving MSE = 0.23 for NS and MSE = 13.1 for CLO, with PSC performance comparable to ophthalmologists while providing consistent quantitative assessments.
• Comparative Analysis: Compared against 14 ophthalmologists and 24 medical students, demonstrating significantly better accuracy for NS and CLO relative to those human readers.
• External Validation: Validated on data from the Singapore Malay Eye Study with cataract severity distributions similar to those in the Age-Related Eye Disease Study (AREDS).

Scientific Applications:

• Clinical assessment: Provides automated quantitative NS, CLO, and PSC measures to support diagnostic evaluation in routine eye examinations.
• Epidemiology and research: Enables large-scale quantitative analyses of cataract prevalence and progression in population and longitudinal studies.

Methodology:

Deep learning models were trained on the AREDS dataset comprising 18,999 photographs from longitudinal follow-up of 1,137 eyes (576 AREDS participants) and externally validated on the Singapore Malay Eye Study.