Circumpapillary OCT-focused hybrid learning for glaucoma grading using tailored prototypical neural networks

Highlights：

• Circumpapillary OCT images in the spectral domain are used for the first time to grade the glaucoma severity.

• The few-shot methodology is redefined in the prototypical paradigm by proposing an optimized k-shot supervised learning which allows exploiting the labelled information.

• Different prototypical-based solutions are conducted and compared with conventional approaches for glaucoma grading.

• The proposed fully supervised prototypical neural network outperforms the previous results achieved in the state-of-the-art for glaucoma detection and glaucoma grading.

• The heat maps extracted are directly in line with the clinician’s opinion, since the highlighted regions of the B-scans correspond to the interesting areas in which the experts focus for glaucoma diagnosis.