Fig. 5
From: OCT-based diagnosis of glaucoma and glaucoma stages using explainable machine learning
(a) Snapshot of the Glaucoma likelihood OCT calculator (GLOC) based on XAI (partial dependency analysis-based predicted probabilities and SHAP-based feature ranking using the RF classifier). (b) Plot of glaucoma likelihood scores versus mean deviation of visual fields. The % values on the right indicate the % of glaucoma cases above that threshold, and the % of normal cases below that threshold. (c) Snapshot of the Glaucoma likelihood OCT web app based on a RF model deployed on the web. (d) Plot of predicted probabilities for glaucoma versus mean deviation of visual fields. The % values on the right indicate the % of glaucoma cases above that threshold and the % of normal cases below that threshold. (e) Correlation between glaucoma likelihood score by the XAI-based Excel app and predicted probability by the RF deployed web app. Demographics: 195 glaucoma eyes (early:86, moderate:72, advanced based on MD:37), 261 normal eyes. The other normal subjects’ MD values were unavailable.
