GLAUDIA: A predicative system for glaucoma diagnosis in mass scanning

Area of Science:

• Ophthalmology
• Medical Imaging
• Artificial Intelligence

Background:

• Glaucoma is a leading cause of vision loss due to retinal ganglion cell (RGC) damage.
• Current diagnostic methods like OCT, SLP, and HRT are costly, limiting widespread glaucoma screening.
• A need exists for an affordable, automated system for early glaucoma detection and mass screening.

Purpose of the Study:

• To develop and evaluate a novel automated glaucoma diagnosis system for mass screening.
• To decrease the false negative rate (FNR) in glaucoma detection.
• To improve early identification of glaucoma to extend symptom-free life.

Main Methods:

• A machine learning model employing a majority voting ensemble of nine techniques was developed.
• The top five highest-accuracy techniques were selected for a consensus ensemble model.
• The models were trained and tested on a dataset of 499 patient records.

Main Results:

• The majority voting model achieved 90% accuracy and an 8% FNR.
• The consensus ensemble model achieved 83% accuracy and a 0% FNR.
• The proposed system demonstrated a dramatic reduction in FNR while maintaining acceptable accuracy.

Conclusions:

• The developed automated glaucoma diagnosis system effectively reduces the FNR.
• The consensus ensemble model shows significant promise for glaucoma mass screening.
• This approach can enhance early detection and management of glaucoma, preserving vision.