FP566

AI based stratification of risk factors in predicting outcomes of DALK in advanced keratoconus

Purpose:To study demographic,imaging &ocular surface risk factors affecting outcomes of DALK in advanced keratoconus(KC).Method:250 eyes with 5 year follow up underwent DALK for advanced KC,were clinically classified as favourable,unfavourable based on graft clarity,suture loosening,vascularisation,vision.Factors including eye rubbing,allergy status,immunoglobulin E(IgE), vitamins D,B12 were assessed.Pre-operative Pentacam(OculusOptikgeräte GmbH)data was exported.AI model was built to assess these factors predicting outcomes of DALK.Result:AI model classified 94.5% &85% cases as favourable or unfavourable respectively based on these risk factors.Systemic allergy,IgE,eye rubbing,Vit D had highest information gain,followed by posterior corneal data from Pentacam.AI model achieved AUC of 0.936.Conclusion:This demonstrates importance of preoperative risk stratification which can affect surgical outcomes using AI.Thus,better control of these factors would enable better management &outcomes.