Indocyanine green angiography and machine learning analysis determine topographic distribution of peripheral hard drusen in a Chinese cohort
摘要
Hard drusen appear as hyperfluorescent dots on indocyanine green angiography (ICGA) due to their high phospholipid content. This study aims to explore the distribution of peripheral hard drusen across age groups and retinal regions, and their correlation with macular pathology including hard drusen, soft drusen, and progression to neovascular age-related macular degeneration (nAMD).
MethodsA hospital-based retrospective study was conducted at Zhongshan Ophthalmic Center. Patients who underwent ICGA over a five-year period were included and categorised into seven age groups. Hard drusen were quantified across nine retinal quadrants using the Trainable Weka Segmentation machine learning plugin. A subgroup of patients underwent imaging follow-up to assess progression to nAMD.
ResultsHere we show, across 1562 normal eyes (14,058 ICGA images) quantified using a validated machine learning model (Dice coefficient: 0.697; sensitivity: 82.93%; ICC: 0.944), that age 50 represents a critical threshold for drusen accumulation, with significant increases observed across all nine quadrants. The nasal quadrants demonstrate earlier and more prominent drusen accumulation compared to the temporal quadrants (P < 0.001 across ages 20–69 years). Notably, the central superior quadrant shows consistent associations with macular hard drusen counts (Stdβ = 0.104, P = 0.010), soft drusen volume (ρ = 0.657, P = 0.020), and progression to nAMD (OR = 1.324, P = 0.040).
ConclusionsPeripheral hard drusen accumulate preferentially in the nasal quadrants and increase markedly after age 50. Comprehensive peripheral retinal evaluation during ICGA examination may hold potential clinical value for AMD monitoring, pending further prospective validation.