Deep Learning-Based Prediction of Individual Geographic Atrophy Progression from a Single Baseline OCT

AI Generated Summary*

A deep learning model forecasting geographic atrophy growth as en face maps from one baseline OCT volume was developed and tested with fivefold cross-validation in 184 eyes of 100 routine-care patients with atrophy secondary to age-related macular degeneration, using manually delineated fundus autofluorescence lesions registered to OCT as reference. Total-lesion Dice was 0.80 at baseline, 0.82 at years one and two, and 0.70 at longer intervals; growth-region Dice ran 0.25 to 0.38, and total-region mean absolute error 0.25 to 0.69 mm. Predicted and manual growth rates correlated moderately (r = 0.61). Areas under the curve for flagging the fastest 10%, 15%, and 20% reached 0.81, 0.79, and 0.77. Single-device data and no external validation limit this pilot work.

*This summary was generated by AI and is published unedited. Oku does not alter these summaries. It may contain errors or omissions and is provided for general informational purposes only. Oku does not guarantee its accuracy, completeness, or reliability. For authoritative information, please refer to the original, peer-reviewed article.