ESCRS - FPS08.06 - An Ai-Based Treatment Advice System, As An App (Imigs), In Making A Smart Personalized Selection Of The Available Minimally Invasive Glaucoma Surgery (Migs) Options

An Ai-Based Treatment Advice System, As An App (Imigs), In Making A Smart Personalized Selection Of The Available Minimally Invasive Glaucoma Surgery (Migs) Options

Published 2022 - 40th Congress of the ESCRS

Reference: FPS08.06 | Type: Free paper | DOI: 10.82333/mtxw-zk52

Authors: Umair Qidwai 1 , gokulan ratnarajan* 1 , owais qidwai 2 , thurka sivapalan 1

1queen victoria hospital,east grinstead,United Kingdom, 2qatar university,doha,Qatar

Purpose

To evaluate the effective predictability of an AI-based Treatment Advice system, as an App (iMIGS), in making a smart personalized selection of the available Minimally Invasive Glaucoma Surgery (MIGS) options, from the baseline clinical parameters. 

Setting

An AI algorithm, Adaptive Neuro Fuzzy Inference System (ANFIS), was used with MIGS data set which was a retrospective case series of patients who underwent either of the four MIGS procedures [Stent, iStent and Endoscopic Cyclophotocoagulation  (ICE2), Preserflo Micro Shunt (PMS) and XEN-45], with or without Phacoemulsification cataract surgery from July 2016 till May 2020 at a single Centre in UK. 

Methods

The AI algorithm was trained using baseline measurements such as age, visual acuity, visual field, intra-ocular pressure (IOP), glaucoma type, number of anti-Glaucoma drops patient is on. The system then predicts one of four possible MIGS which would be most suitable for the patient.

Results

The proposed ANFIS system was found to be 91% accurate (89% on average) with high Sensitivity (80%) and Specificity (90%), as well as low false positive rate (7%) and low miss-rate (20%). 

Conclusions

The ANFIS technique outperformed the other popular techniques (such as Support Vector Machine, Shallow Neural Networks and Regression ,models) with a multitude of scales and has shown very promising approach for predicting the treatment classes based on the initial clinical measurements.