Automatically grading keratoconus progression using a novel decision support software tool: an agreement analysis

Abstract

Purpose

To present and clinically validate a novel software tool that automatically compares tomographic parameters and evaluates keratoconus progression based on prespecified criteria.

Materials and Methods

Two graders (one cornea specialist and one general ophthalmologist) evaluated two tomographic scans (Pentacam; Oculus Optikgeräte GmbH, Wetzlar, Germany) of keratoconus patients separated by more than 4 months and subjectively classified each image pair in 0 – no progression, 1 – doubtful progression, 2 – clear progression. The agreement between the automated software and each grader was evaluated: percentage of agreement and Cohen’s kappa coefficient (k): 0 to 0.2 (poor); 0.2 to 0.4 (fair); 0.4 to 0.6 (moderate); 0.6 to 0.8 (substantial) and 0.8 to 1.0 (almost perfect).

Results

We included 43 eyes from 26 keratoconus patients (age 27.9±6.9 years) that underwent consecutive tomographic evaluation (7.5±2.6 months between exams). The agreement between the software and a cornea specialist was very high, 92.3% [95% CI 0.88-0.97, p<0.001], with a Cohen’s K of 0.80 [95% CI 0.67-0.93, p<0.001], representing almost perfect agreement. The agreement with a general ophthalmologist was lower, 88.4% [95% CI 0.83-0.93, p<0.001], with a Cohen’s K of 0.66 [95% CI 0.51-0.83, p<0.001], representing substantial agreement.

Conclusions

The proposed software tool demonstrates excellent agreement with an expert grader in keratoconus progression. This tool may be useful in quickly comparing critical tomographical parameters in serial exams of keratoconus patients, thus potentially improving clinical workflows. Furthermore, it may be a useful adjunct to the general ophthalmologist who may not be familiar with either tomographic examinations and/or evaluation of keratoconus progression.