Ethnicity optimized indices enhance the diagnostic efficiency of early Keratoconus: A multicenter validation study.

Abstract

Purpose: This study aimed to evaluate the effectiveness of an ethnically-optimized corneal tomography and biomechanics index using artificial intelligence (AI) techniques for diagnosing early ectasia in Chinese populations and determine the diagnostic indices' sensitivity, specificity, and cutoff value for clinical practice.

Methods: This multicenter case-control retrospective study included 1,012 eyes from three centers in China (Tianjin, Shaanxi, and Shandong). The groups included very asymmetric ectasia with normal topography (VAE-NT, n = 146), contralateral ectasia (VAE-E, n = 127), bilateral keratoconus (KC, n = 247), and normal eyes (NL, n = 492). The diagnostic efficiency of the Corvis Biomechanical Index (CBI), CBI for Chinese populations (cCBI), Tomographic and Biomechanical Index version 1 (TBIv1), TBI version 2 (TBIv2), and TBI for Chinese populations (cTBI) was assessed using receiver operating characteristic (ROC) curves. The diagnostic efficiency was compared using DeLong's test.

Results: cTBI had the highest diagnostic accuracy for distinguishing NL from early ectasia (VAE-NT), with an area under the ROC (AUROC) of 0.93 at the cutoff value of 0.41, a sensitivity of 84.39 % and specificity of 92.67 %. cTBI diagnostic efficacy in early ectasia was better than TBIv2, TBIv1, cCBI, and CBI (Delong test, P < 0.01). It also had the best efficacy in distinguishing NL from "disease" (VAE-NT + VAE-E + KC), with an AUROC of 0.98 at the cutoff value of 0.66, sensitivity of 91.05 %, and specificity of 99.39 %. cTBI achieved 100 % sensitivity and specificity for distinguishing NL from clinical ectasia (VAE-E + KC) at a cutoff of 0.81.

Conclusions: The ethnicity optimized cTBI index had good diagnostic efficiency for early ectasia in the Chinese population, highlighting the benefit of ethnic-specific parameter optimization.