Dr. HARSHAVARDHAN KAVALADANDI

Dr.MEENA MENON, Dr. PRIYANKA SUDHAKAR, Dr.SHREESHRUTHI N

Abstract

AIM:

To validate the ability of Artificial Intelligence (AI) based algorithm to assess cup disc ratio (CDR), ISNT rule and Disc Damage Likelihood Scale (DDLS) in Optic Disc evaluation for glaucoma.

METHODS:

A total of 120 patients visiting the glaucoma clinic were enrolled for the study. Dilated 45-degree disc centred fundus photographs were captured using a fundus camera integrated with Netra.AI deep learning (www.leben.ai).The results obtained by AI were compared with that of Glaucoma experts.

RESULTS:

Statistical analysis was performed between AI results and human grader. The agreement across parameters was found between 0.33 to 0.61 kappa. Sensitivity & Specificity of AI in detecting the likelihood of glaucomatous disc were found to be >75%.

CONCLUSION:

A deep learning algorithm is a promising tool in quick and accurate disc assessment. It has a great potential to be an excellent screening test at primary care levels thereby helping in early detection and management of glaucoma.

Full Text