A Workflow for Computer-Aided Diagnosis of Glaucoma

Glaucoma is an optic disease with visual impairment, leading to blindness. The timely diagnosis of glaucoma can reserve the blinding process with a good prognosis. Although typical glaucoma is characterized by elevated intraocular pressure and optic neuropathy, accurate diagnosis is still challenging by measuring intraocular pressure and ophthalmoscopy diagnosis. This paper presents a workflow to diagnose glaucoma and distinguish low-quality fundus images automatically. The performance in the hidden test set shows the promising clinical usefulness against experts, which achieved first place in the AIROGS challenge with the partial AUC (90%~100% specificity) of 0.8998 and sensitivity of 0.8533 at 95% specificity. The algorithm is available at https://grand-challenge.org/algorithms/airogs_fine3/.