Optic disc drusen affect 145 million people worldwide, causing unpredictable and irreversible vision loss in up to 87% of patients[1]. Sometimes presenting in childhood, these calcified deposits compress optic nerve fibers, necessitating close clinical monitoring[2]. Yet, the clinical course of drusen is highly variable, and clinicians lack reliable predictors of vision loss, leaving patients uncertain about prognosis and dependent on years of repeated imaging. Growing evidence suggests that vision loss may be driven by interactions between drusen morphology (size, location), and patient factors (age, vascular anatomy, compliance)[3]. Optical coherence tomography (OCT) provides non-invasive, near-cellular resolution 3D imaging of the eye and is now the standard for detecting and monitoring drusen[4]. However, the scale and complexity of OCT datasets make manual analysis impractical in routine care, leaving a critical unmet need for automated approaches that can transform imaging data into clinically meaningful prognostic information[5]. In this project, we will analyze previously collected OCT imaging and clinical data from 400 patients with drusen to determine long-term predictors of vision loss. We will develop an explainable artificial intelligence (AI) approach that (i) applies a validated model to automatically detect and characterize drusen, and (ii) builds a predictive model integrating ocular and patient features to predict vision loss at 5 years. By identifying the structural and biological drivers of drusen progression, this work will use explainable AI to shift monitoring from reactive surveillance to predictive, personalized care – enabling early identification of high-risk patients and more targeted management before irreversible vision loss occurs