Mathew Francis1, Natasha Pahuja1, Rushad Shroff1, Roshan Gowda1, Himanshu Matalia1, Rohit Shetty1, Everette J Remington Nelson1, Abhijit Sinha Roy2. 1. From the Imaging, Biomechanics and Mathematical Modeling Solutions (Francis, Sinha Roy), Narayana Nethralaya Foundation, the Corneal and Refractive Surgery Division (Pahuja, Shroff, Gowda, Matalia, Shetty), Narayana Nethralaya, Bangalore, and the School of Biosciences and Technology (Nelson), Vellore Institute of Technology, Vellore, India. 2. From the Imaging, Biomechanics and Mathematical Modeling Solutions (Francis, Sinha Roy), Narayana Nethralaya Foundation, the Corneal and Refractive Surgery Division (Pahuja, Shroff, Gowda, Matalia, Shetty), Narayana Nethralaya, Bangalore, and the School of Biosciences and Technology (Nelson), Vellore Institute of Technology, Vellore, India. Electronic address: asroy27@yahoo.com.
Abstract
PURPOSE: To evaluate the performance of waveform-derived variables in distinguishing normal, suspect, and keratoconic eyes. SETTING: Narayana Nethralaya Eye Hospital, Bangalore, India. DESIGN: Retrospective case series. METHODS: Scheimpflug tomography (Pentacam) and dynamic Scheimpflug analysis (Corvis ST) of 253 normal (253 patients) eyes and 205 keratoconic eyes (205 patients) were evaluated. Among the 205 patients, 62 had keratoconus in 1 eye, while the unaffected eye was suspect. From deformation amplitude, deflection amplitude and whole-eye movement were extracted. A biomechanical model was used to derive a linear (kc [constant]) and nonlinear measure (kc [mean]) of corneal stiffness. Multivariate logistic regression was performed to determine sensitivity and specificity. The analysis was validated in another dataset of 59 normal, 45 suspect, and 160 keratoconic eyes. RESULTS: Deformation amplitude maximum, applanation 1 time and deformation amplitude, applanation 2 time, kc (constant), kc (mean), and deflection amplitude maximum were significantly different between normal and keratoconic eyes (P < .001). The deformation characteristics of the suspect eyes were similar to those of the keratoconic eyes, particularly grade 1 (P > .05). The kc (constant) and kc (mean) had the highest area under curve (>0.98), sensitivity, and specificity greater than 90% and 91%, respectively. Logistic regression using kc (constant) and kc (mean) improved the area to 1.0, with a sensitivity and specificity equal to 99.6% and 100%, respectively. In the validation dataset, the same cutoff yielded a sensitivity, specificity, and accuracy of 99.5%, 100%, and 99.6%, respectively. CONCLUSION: Corneal stiffness and waveform analyses could be reliable differentiators of suspect and keratoconic eyes from normal eyes.
PURPOSE: To evaluate the performance of waveform-derived variables in distinguishing normal, suspect, and keratoconic eyes. SETTING: Narayana Nethralaya Eye Hospital, Bangalore, India. DESIGN: Retrospective case series. METHODS: Scheimpflug tomography (Pentacam) and dynamic Scheimpflug analysis (Corvis ST) of 253 normal (253 patients) eyes and 205 keratoconic eyes (205 patients) were evaluated. Among the 205 patients, 62 had keratoconus in 1 eye, while the unaffected eye was suspect. From deformation amplitude, deflection amplitude and whole-eye movement were extracted. A biomechanical model was used to derive a linear (kc [constant]) and nonlinear measure (kc [mean]) of corneal stiffness. Multivariate logistic regression was performed to determine sensitivity and specificity. The analysis was validated in another dataset of 59 normal, 45 suspect, and 160 keratoconic eyes. RESULTS: Deformation amplitude maximum, applanation 1 time and deformation amplitude, applanation 2 time, kc (constant), kc (mean), and deflection amplitude maximum were significantly different between normal and keratoconic eyes (P < .001). The deformation characteristics of the suspect eyes were similar to those of the keratoconic eyes, particularly grade 1 (P > .05). The kc (constant) and kc (mean) had the highest area under curve (>0.98), sensitivity, and specificity greater than 90% and 91%, respectively. Logistic regression using kc (constant) and kc (mean) improved the area to 1.0, with a sensitivity and specificity equal to 99.6% and 100%, respectively. In the validation dataset, the same cutoff yielded a sensitivity, specificity, and accuracy of 99.5%, 100%, and 99.6%, respectively. CONCLUSION:Corneal stiffness and waveform analyses could be reliable differentiators of suspect and keratoconic eyes from normal eyes.
Authors: Louise P G Esporcatte; Marcella Q Salomão; Nelson S Junior; Aydano P Machado; Érica Ferreira; Tomás Loureiro; Renato A Junior Journal: Saudi J Ophthalmol Date: 2022-07-11