Development of a classification system based on corneal biomechanical properties using artificial intelligence predicting keratoconus severity

Abstract

Background To investigate machine-learning (ML) algorithms to differentiate corneal biomechanical properties between different topographical stages of keratoconus (KC) by dynamic Scheimpflug tonometry (CST, Corvis ST, Oculus, Wetzlar, Germany). In the following, ML models were used to predict the severity in a training and validation dataset. Methods Three hundred and eighteen keratoconic and one hundred sixteen healthy eyes were included in this monocentric and cross-sectional pilot study. Dynamic corneal response (DCR) and corneal thickness related (pachymetric) parameters from CST were chosen by appropriated selection techniques to develop a ML algorithm. The stage of KC was determined by the topographical keratoconus classification system (TKC, Pentacam, Oculus). Patients who were classified as TKC 1, TKC 2 and TKC 3 were assigned to subgroup mild, moderate, and advanced KC. If patients were classified as TKC 1–2, TKC 2–3 or TKC 3–4, they were assigned to subgroups according to the normative range of further corneal indices (index of surface variance, keratoconus index and minimum radius). Patients classified as TKC 4 were not included in this study due to the limited amount of cases. Linear discriminant analysis (LDA) and random forest (RF) algorithms were used to develop the classification models. Data were divided into training (70% of cases) and validation (30% of cases) datasets. Results LDA model predicted healthy, mild, moderate, and advanced KC eyes with a sensitivity (Sn)/specificity (Sp) of 82%/97%, 73%/81%, 62%/83% and 68%/95% from a validation dataset, respectively. For the RF model, a Sn/Sp of 91%/94%, 80%/90%, 63%/87%, 72%/95% could be reached for predicting healthy, mild, moderate, and advanced KC eyes, respectively. The overall accuracy of LDA and RF was 71% and 78%, respectively. The accuracy for KC detection including all subgroups of KC severity was 93% in both models. Conclusion The RF model showed good accuracy in predicting healthy eyes and various stages of KC. The accuracy was superior with respect to the LDA model. The clinical importance of the models is that the standalone dynamic Scheimpflug tonometry is able to predict the severity of KC without having the keratometric data. Trial registration NCT04251143 at Clinicaltrials.gov, registered at 12 March 2018 (Retrospectively registered).

Distribution of pediatric keratoconus by different age and gender groups

“…Corneal topography and tomography were obtained through Pentacam HR (Oculus, Wetzlar, Germany), which uses the Scheimpflug imaging technique to present indices with acceptable accuracy and repeatability ( 23 ). The following parameters were evaluated in the current study: corneal thickness (apex corneal thickness: ACT; pupil corneal thickness: PCT; thinnest corneal thickness: TCT), keratometry (flat keratometry: K1; steep keratometry: K2; mean keratometry: Km; maximum keratometry: Kmax), inferior superior value (IS-value), BAD, and TKC (a classification system of keratoconus as provided by Pentacam HR) ( 24 ).…”

Artificial Intelligence–Based Diagnostic Model for Detecting Keratoconus Using Videos of Corneal Force Deformation

“… 21 Likewise, other studies reported good KC detection ability using corneal dynamic response parameters from the Corvis ST with linear regression models and random forests. 13 , 15 It is noteworthy that these parameters are directly generated by the Ocular Response Analyzer or Corvis ST, and the dependency of the diagnostic model on the device may affect the general applicability of the model. Furthermore, it has also been demonstrated that, if there is a correlation between parameters, both the accuracy and consistency of the model predictions are compromised.…”

“… 11 – 13 Moreover, previous studies have found correlations only between KC and certain biomechanical indices, and to date, no diagnostic model has been found to directly diagnose KC based on corneal biomechanical alterations. 14 , 15 The current study, with the aid of ML approaches, attempts to analyze the corneal dynamic deformation process to diagnose KC solely from a corneal biomechanical perspective.…”

The role of corneal biomechanical properties assessment in laser vision correction – the introduction

“…The deformation amplitude (DA) is defined as the greatest inward dislocation of the corneal apex measured at the highest concavity (HC) point [11]. DA ratio of central and peripheral deflection is assessed in a distance of 1 mm and 2 mm resulting in DAR-1 and DAR-2 parameters and stiffness parameter at the 1 st applanation (SPA1) [12]. Applanation lengths (AL) and corneal velocities (CVel) are measured in inward and outward phases.…”

“…The curvature radius at the highest concavity (curvature radius HC) integrated inverse radius (IntInverseR) and maximum inverse radius (InverseR) are also documented. The higher values of IR, the lower corneal resistance to deformation what indicate lower corneal stiffness [12]. Corneal pachymetry (Pachy) and parameters essential in potential keratoconus detection such as Ambrosio Rational Thickness horizontal (ARTh) and Pachyslope are measured before the air-puff generation [12].…”