Analysis of corneal biomechanics using ultra high-speed Scheimpflug imaging to distinguish normal from keratoconic patients

Abstract

 

Purpose

 

To evaluate the ability of a high-speed Scheimpflug camera (Corvis ST,Oculus,Wetzlar,Germany) to distinguish between normal and keratoconic(kc)eyes,by comparing new corneal deformation parameters (CDP).

 

Methods

 

Retrospective comparative study including 792 eyes, a normal group comprised of 587 eyes and 205 keratoconic cases. Diagnosis was based on clinical examinations, including Placido-disk based corneal topography and rotating Scheimpflug corneal tomography. Data from the Corvis ST were extracted using research software including Intraocular pressure (IOP), central corneal thickness (CCT) and a total of 56 CDP. Additionally, a new IOP correction (IOPc) algorithm based on Ocular Biomechanics Group of the University of Liverpool (UK) that includes CDPs, CCT and age. Logistic regression technique was used to evaluate the discriminant faculty of each parameter and then receiver operating characteristic (ROC) curves were calculated. The independent parameter with an area under the receiver operating characteristic curve (AUC) over 0.7 and a sensibility over 0.75 with a specificity non less than 0.5 were included in a multivariable logistic regression to combine parameters in order to provide best possible separation between normal patients and kc.

 

Results

 

Statistically significant differences between kc and normal eyes were found in all CDPs (p<0.05) except for 10 non significant CDPs that were excluded. There were 3 CDP with AUC higher than 0.8. The best individual CDP was ratio between deformation amplitude and deformation amplitude at 2 mm (DA/DA2) with AUC of 0.8680.<br /> The CDPs included in the multivariable logistic regression were second applanation time, first applanation deflection amplitude, maximum deflection amplitude, slope of deflection amplitude, frame of second applanation, inverse concave radius, simulated time of second applanation and DA/DA2. The analysis was also adjusted for age and IOPc.<br /> The multivariate logistic regression revealed an AUC of 0.9938, providing a very high predictive accuracy.

 

Conclusions

 

Our study demonstrates that Corvis ST is able to provide data that is able to distinguish between normal and keratoconic patients with high predictive accuracy by combining deformation parameters. The integration of CDPs from Corvis ST and tomographic data is promising to further enhance the screening accuracy for ectatic diseases.