A deep learning algorithm was developed to predict individual geographic atrophy (GA) progression from a single baseline optical coherence tomography (OCT) scan. The study evaluated the algorithm's ability to predict GA growth over 3 years and identify fast progressors. The algorithm used a deep learning-based method to model GA lesion growth over time from a single baseline OCT scan. It was validated using 184 eyes from 100 patients with GA secondary to age-related macular degeneration (AMD). Fundus autofluorescence (FAF) images were manually annotated and anatomically registered to OCT scans to create 2D en face OCT annotations, which served as a reference for model performance. The algorithm achieved a mean Dice similarity coefficient (DSC) of 0.80 for the total GA region at baseline, increasing to 0.82 over the first 2 years, and slightly decreasing to 0.70 after 3 years. The mean absolute error (MAE) was low over the first year and slowly increased over time, ranging from 0.25 mm to 0.69 mm for the total GA region. The algorithm achieved an area under the curve (AUC) of 0.81, 0.79, and 0.77 for identifying the top 10%, 15%, and 20% GA growth rates, respectively. The algorithm was able to reliably predict GA growth and identify fast progressors, providing interpretable outputs in the form of en face maps. The results suggest that the algorithm could be a valuable tool for clinical decision support in managing GA, particularly with the availability of new treatments. The study highlights the potential of deep learning in predicting GA progression and identifying high-risk patients for targeted interventions. The algorithm was validated using a fivefold cross-validation approach, with performance metrics including DSC and MAE. The study also discusses the limitations of the algorithm, including potential selection bias and the need for further validation on different OCT devices. The algorithm's ability to predict GA growth and identify fast progressors could improve patient management and treatment decisions in GA. The study emphasizes the importance of accurate GA progression prediction for personalized treatment strategies. The algorithm's performance was evaluated using a combination of DSC and MAE metrics, with results showing good agreement between manual annotations and model predictions. The study concludes that the algorithm is capable of fully automated GA lesion growth prediction from a single baseline OCT scan, providing a promising step toward clinical decision support tools for GA management.A deep learning algorithm was developed to predict individual geographic atrophy (GA) progression from a single baseline optical coherence tomography (OCT) scan. The study evaluated the algorithm's ability to predict GA growth over 3 years and identify fast progressors. The algorithm used a deep learning-based method to model GA lesion growth over time from a single baseline OCT scan. It was validated using 184 eyes from 100 patients with GA secondary to age-related macular degeneration (AMD). Fundus autofluorescence (FAF) images were manually annotated and anatomically registered to OCT scans to create 2D en face OCT annotations, which served as a reference for model performance. The algorithm achieved a mean Dice similarity coefficient (DSC) of 0.80 for the total GA region at baseline, increasing to 0.82 over the first 2 years, and slightly decreasing to 0.70 after 3 years. The mean absolute error (MAE) was low over the first year and slowly increased over time, ranging from 0.25 mm to 0.69 mm for the total GA region. The algorithm achieved an area under the curve (AUC) of 0.81, 0.79, and 0.77 for identifying the top 10%, 15%, and 20% GA growth rates, respectively. The algorithm was able to reliably predict GA growth and identify fast progressors, providing interpretable outputs in the form of en face maps. The results suggest that the algorithm could be a valuable tool for clinical decision support in managing GA, particularly with the availability of new treatments. The study highlights the potential of deep learning in predicting GA progression and identifying high-risk patients for targeted interventions. The algorithm was validated using a fivefold cross-validation approach, with performance metrics including DSC and MAE. The study also discusses the limitations of the algorithm, including potential selection bias and the need for further validation on different OCT devices. The algorithm's ability to predict GA growth and identify fast progressors could improve patient management and treatment decisions in GA. The study emphasizes the importance of accurate GA progression prediction for personalized treatment strategies. The algorithm's performance was evaluated using a combination of DSC and MAE metrics, with results showing good agreement between manual annotations and model predictions. The study concludes that the algorithm is capable of fully automated GA lesion growth prediction from a single baseline OCT scan, providing a promising step toward clinical decision support tools for GA management.