Integration of Corneal Tomography and Biomechanical Parameters for Diagnosis of Ectatic Disease

Ambrósio, Renato

Integration of Corneal Tomography and Biomechanical Parameters for Diagnosis of Ectatic Disease

Friday, April 17, 2015

KIOSKS (San Diego Convention Center)

Renato Ambrósio Jr, MD, PhD

Allan Luz, MD

Bernardo T. Lopes, MD

Rosane Correa, MD

Ana Laura C. Canedo, MD

Bruno F. Valbon, MD

Isaac C. Ramos, MD

João Marcelo Lyra, MD

Purpose
To investigate the accuracy of the integration of corneal biomechanical parameters from Corvis ST for detecting ectatic corneal disease and to test if these metrics provide benefit when combined to Pentacam tomographic (3D elevation, curvature data and thickness distribution) data for detecting ectatic corneal disease.

Methods
Methods: 475 normal eyes (N), 248 eyes with keratoconus (KC) and = 66 cases of forme fruste keratoconus (FFKC) were retrospectively reviewed. Corneal deformation data was obtained from dynamic ultra-high speed Scheimpflug imaging (Corvis ST, Oculus, Wetzlar – Germany). Tomographic indices were obtained using the Oculus Pentacam HR (Wetzlar, Germany). The eyes were classified accordingly to clinical data. Criteria for diagnosis of keratoconus was accomplished accordingly to the Collaborative Longitudinal Evaluation of Keratoconus (CLEK) Study. FFKC criteria was the eye with no clinical or topographic evidence of keratoconus, from patients with keratoconus diagnosed in the fellow eye. The ability of the parameters to distinguish normal (N) and ectatic cases (FFKC and KC) was assessed by receiver operating characteristic (ROC) curve analysis. Linear regression analysis was accomplished to optimize accuracy.

Results
Area under the ROC curve (AUC) higher than lower than 0.79 for all Corvis ST derived parameters. The best parameter was BAD-D with AUC of 0. 978.

The combination of tomographic parameters and Corvis ST data significantly enhanced the ability to separate normal and ectatic cases. We present two functions which were calculated using different strategies of logistic regression analysis with AUC of 0.986 and 0.991.

Conclusion
Corneal biomechanics significantly improved the ability of tomographic data for detecting ectasia. This seems to be more relevant when considering milder forms of ectasia such as the cases with FFKC. Novel artificial intelligence functions should be included. Enhcanced clinical correlations are also to be performed. Some of the cases of FFKC may truly be unilateral cases of ectasia.