Gradient-Boosting Classifiers Combining Vessel Density and Tissue Thickness Measurements for Classifying Early to...

Area of Science:

• Ophthalmology
• Medical Imaging
• Machine Learning

Background:

• Glaucoma diagnosis relies on detecting structural and functional changes.
• Optical coherence tomography (OCT) and OCT angiography (OCTA) provide detailed imaging of the optic nerve head (ONH) and macula.
• Current diagnostic tools have limitations in early glaucoma detection.

Purpose of the Study:

• To compare gradient-boosting classifier (GBC) analysis of OCTA-measured vessel density (VD) and OCT-measured tissue thickness with standard OCTA/OCT parameters for classifying healthy versus glaucomatous eyes.
• To evaluate the diagnostic accuracy of different GBC models combining various OCTA and OCT measurements.

Main Methods:

• 180 healthy and 193 glaucomatous eyes underwent OCTA and OCT imaging.
• Four GBCs were developed: Macula GBC, ONH GBC, vessel density GBC, and thickness GBC.
• A full GBC combining all parameters was also evaluated, with diagnostic accuracy compared to standard parameters using ROC curve analysis (AUROC).

Main Results:

• GBCs demonstrated superior diagnostic accuracy (AUROCs 0.90–0.93) compared to standard OCTA/OCT parameters (AUROCs 0.64–0.91).
• The full GBC achieved the highest AUROC of 0.93.
• Macular thickness and ONH VD measurements were identified as the most influential parameters in the full GBC model.

Conclusions:

• GBCs integrating OCTA and OCT macula and ONH measurements enhance glaucoma detection accuracy.
• This approach shows promise in improving upon existing instrument-provided diagnostic parameters.
• Further validation is needed, but GBCs offer a powerful tool for early glaucoma identification.