Abstract
Keratoconus (KC) is a contraindication for corneal refractive surgery and can cause serious damage to the patient's vision. To improve the safety of the surgery, we aim to distinguish between suspected KC and subclinical KC, while also developing a predictive model for KC. This cross-sectional study investigated 116 eyes with suspected KC (I-S Value > 1.4 D and ≤ 1.9 D) and 28 eyes with subclinical KC. All the Chinese subjects were examined via the Pentacam and Corvis ST (Oculus, Wetzlar, Germany). The 90 parameters of the Belin/Ambrósio System, Topometric/KC System, and Corvis ST System were extracted from internal CSV files and analyzed. The differences in all the parameters between the two groups were compared, and ROC curves were created. LASSO regression was employed to simplify the number of parameters, and a logistic regression model was constructed for KC prediction. The suspected KC group, in contrast to the subclinical KC group, predominantly consisted of females and older patients. The highest AUC parameters of the three systems were PachyProgIndexMax [0.896 (0.839, 0.954), cutoff = 1.445], IHD [0.787 (0.691, 0.883), cutoff = 0.029], SPA1 [0.811 (0.718, 0.904), cutoff = 87.485]. For the first discovery, the corneal diameter in the suspected KC group (11.56 ± 0.38) was smaller than that in the subclinical KC group (11.91 ± 0.33) (P < 0.05) [AUC = 0.736 (0.641, 0.832), cutoff = 11.825]. Seven relevant parameters were identified via LASSO regression (AUC = 0.954), including IHD, Cornea Diameter, DARatioMax_2 mm, BADDy, BADD, PachyProgIndexMax, and CBI. The prediction accuracy of the logistic regression model was 0.902 (AUC = 0.964). Our model effectively predicts an elevated risk of suspected KC to subclinical KC and even KC in younger Chinese men, particularly those exhibiting increased corneal diameter and morphological and biomechanical parameters. This model might help with screening for preoperative refractive surgery, thereby improving surgical safety.