Predicting Keratoconus Progression Using High Resolution Oct Measurement And A Novel Artificial Intelligence Model

Published 2024 - 42nd Congress of the ESCRS

Reference: PP01.04 | Type: Free paper | DOI: 10.82333/dkv4-3m05

Authors: Andreas Honeder* ¹ , Leon Pomberger ¹ , Klemens Waser ¹ , Haidar Khalil ¹ , Matthias Bolz ¹ , Peter Laubichler ¹ , Nino Hirnschall ¹

¹Ophthalmology and Optometry,Kepler University Clinic, Johannes Kepler University,Linz,Austria

Purpose

Aim of this study was to predict the progression of keratoconus using a novel artificial intelligence prediction model trained with data of 12 month follow ups.

Setting

Kepler University Clinic Linz, MCIII, Department of Ophthalmology and Optometry and Johannes Kepler University.

Methods

Patient data was collected from our Keratoconus outpatient clinic in a retrospective fashion. In all cases a Scheimpflug measurement device (Pentacam HR, Oculus, Germany) and a spectral domain OCT (MS-39, CSO, Italy) were available at least over a time period of 12 months. Progression was defined as if either Kmax increased in ≥1 Diopter or thinnest pachymetry had a change of at least -10µm within 12 months. Bayes statistic was used to find the best predictive parameters plus a machine learning approach to create an image based AI model.

Results

In total 345 eyes of 173 patients were included in this study. Mean age was 33.13 years (SD 13.56; median: 30; range 8 to 74). Preliminary results showed a Kmax of 52.47 D (SD: 7.85) and a mean tomographic astigmatism of 3.08 D (SD: 2.12). Thinnest pachymetry was on average 470.64µm (SD: 65.10). Of the different models the Bayesian progression model was found to be promising as well as a random forest machine learning approach.

Conclusions

Prediction of keratoconus progression using Bayesian statistics is a powerful tool that could be used additionally to “conventional” machine learning concepts.