Abstract
BACKGROUND: Age-related macular degeneration is a leading cause of central vision loss, and assessing visual function with microperimetry can be time-consuming and tiring for patients. Targeting regions corresponding to worsening acute-stage retinal lesions may reduce test durations and patient fatigue. RESULTS: We developed a machine-learning approach using multi-modal imaging data to differentiate lesional regions from healthy retinal areas. Our dataset included 344,003 regions extracted from color fundus photographs, infrared fundus images, optical coherence tomography, and optical coherence tomography angiography images. A gradient-boosted tree-ensemble model was trained on this data and achieved an area under the receiver operating characteristic curve of 0.95 in detecting end-stage lesions in chronic age-related macular degeneration. CONCLUSIONS: The proposed method effectively detects lesions associated with age-related macular degeneration using multi-modal imaging and machine learning. This approach offers a potential solution for creating targeted microperimetry test patterns, which can reduce testing time and patient fatigue, thereby enhancing the clinical assessment of visual function in affected patients.