AS-OCT provides a non-invasive, non-contact imaging method in the field of ophthalmology. Its rapid scan speed minimizes the effects of eye-movements and has an advantage over some Scheimpflug-based equipment. Reinstein et al. [6, 7, 15] have described the epithelial compensatory mechanism due to stromal irregularities in KC, providing a perspective on the detection of this corneal ectatic disease. Several recent studies have utilized the OCT technology to distinguish KC from healthy corneas by analyzing the corneal epithelium thickness profile [5, 8, 10, 22,23,24,25,26]. The results of this study showed that FFKC and KC had higher epithelial aberrations compared to normal eyes, several parameters exhibited the ability for diagnosing FFKC and KC.

To identify suitable candidates for LVC surgery, it is crucial to screen for early-stage KC during preoperative assessments, given that it is a major risk factor for post-LVC ectasia [27, 28]. Previous studies had attempted to diagnose FFKC using various metrics such as corneal thickness, topographical, biomechanical, or aberrations data, yet this remains challenging for ophthalmologists. Epithelial thickness maps may help identify FFKC in patients who have an otherwise normal topography. Temstet et al. reported that epithelial thickness in the thinnest corneal zone had 0.79 of AUC in discrimination of FFKC [29]. However, in Toprak et al.’s study, this parameter showed no statistically significant ability for distinguishing FFKC. Thinnest pachymetry and stromal thicknesses [center, nasal, superior, inferior, inferior-nasal (IN), superior-nasal (SN), inferior-temporal (IT), superior-temporal (ST), and at the thinnest epithelial point] in discrimination of eyes with FFKC from normal controls (AUC ranges between 0.614 and 0.712) [10]. Similar results were also obtained in our study. Supplemental Table 2 presents the diagnostic performance of epithelial thickness across different corneal regions, with AUC values for FFKC ranging from 0.532 to 0.717. Supplemental Table 3 presents the diagnostic performance of stromal thickness measurements, with AUC values ranging from 0.660 to 0.755 for FFKC. Koh et al. [30] conducted a study measuring ocular and corneal HOAs in pre-topographic KC (equivalent to 'FFKC' in our study) and evaluated the ability of wavefront data to distinguish FFKC from normal eyes. Their results showed all ocular and corneal wavefront parameters were not different between FFKC and control groups, whereas our study showed that all epithelial wavefront parameters (total RMS, HOAs RMS, coma, trefoil, SA, Ast II) were significantly higher in the FFKC group. This discrepancy suggests that while the ocular and total cornea remain optically unchanged, the epithelial wavefront analysis could detect variations in the pathological cornea from the earliest stage of KC progression. In the diagnosis of FFKC, Koh’s study [30] indicated that ocular wavefront data were insufficient for detecting FFKC, but the corneal HOAs RMS (AUC: 0.781, sensitivity: 100.0%, specificity: 47.0%) and corneal coma (AUC: 0.735, sensitivity: 73.0%, specificity: 73.0%) exhibited a potential to discriminate between FFKC and control eyes. Saad et al. [31] indicated that corneal coma had an AUC of 0.778 with 71.0% sensitivity and 80.0% specificity. According to our data, the epithelial HOAs RMS (AUC: 0.714, sensitivity: 80.5%, specificity: 54.9%) and epithelial coma (AUC: 0.788, sensitivity: 81.6%, specificity: 62.6%) demonstrated high discriminatory ability and also had the potential to differentiate FFKC from healthy corneas. To optimize diagnostic accuracy, we employed stepwise logistic regression to construct the EAI (EAI-FFKC), a discriminant function for distinguishing FFKC from healthy corneas. The EAI-FFKC showed the best discriminatory power by AUC value of 0.822 with 77.0% sensitivity and 75.8% specificity. This multi-parameter diagnostic approach, combining epithelial HOAs RMS, coma, and SA, demonstrated greater diagnostic accuracy than any single parameter in distinguishing FFKC from normal corneas, improving both sensitivity and specificity. Previous studies investigated corneal wavefront indices generated from Scheimpflug, Placido, and Hartmann–Shack based devices with acceptable validity for differentiating normal corneas from early subclinical keratoconus (SKCN). The front Baiocchi-Calossi-Versaci (BCV) index from Sirius was the most accurate parameter for diagnosing SKCN (AUC = 0.887), followed by vertical coma (AUC = 0.857) with Pentacam and OPD-Scan III (AUC = 0.857) [32]. An eye was diagnosed with SKCN if it had at least three of the following abnormal topographic criteria: a skewed asymmetric bow tie, central or inferior steepening in anterior/posterior elevation height. SKCN is the later form of FFKC, so there is a slightly higher AUC value [33, 34].

Supplemental Table 1 shows the diagnostic performance of total corneal aberrations. In the FFKC group, the AUC values for epithelial HOAs RMS and coma were consistently higher than those of total corneal aberrations (epithelial vs. total: HOAs RMS, 0.714 vs. 0.701; coma, 0.788 vs. 0.728). In contrast, epithelial trefoil and Ast II yielded lower AUCs compared to total corneal aberrations (epithelial vs. total: trefoil, 0.605 vs. 0.680; Ast II, 0.651 vs. 0.680). A similar trend was observed in the mild KC group (HOAs RMS, 0.974 vs. 0.957; coma, 0.976 vs. 0.958; trefoil, 0.899 vs. 0.951; Ast II, 0.890 vs. 0.952). Additionally, our comparison between FFKC and mild KC groups provided further insight into early KC progression. The AUC values of all parameters were higher when distinguishing between FFKC and mild KC than between FFKC and normal corneas, suggesting that FFKC eyes more closely resemble healthy eyes than KC eyes.

Mohammadpour et al. found that the natural shape of the cornea changes over time, causing visual distortions and increasing aberrations [35]. Salman et al. believe that analyzing the different characteristics of aberrations can improve diagnostic accuracy, especially in the early stages of KC [36]. Previous studies [30, 36,37,38,39] reported that higher levels of ocular and corneal aberrations were exhibited in clinical KC and topographic KC compared to normal eyes. Our data also showed that all epithelial wavefront parameters of KC were significantly higher than normal eyes. Interestingly, as KC progressed through stages from FFKC to mild, moderate, and severe, there were corresponding increases in all types of epithelial aberrations. This revealed that the epithelial compensation mechanism was presented throughout the entire disease progression of KC. Concerning KC diagnosis, Saad et al. [31] reported that corneal coma and trefoil reached AUC values of 0.988 (sensitivity: 98.0%, specificity: 99.0%) and 0.960 (sensitivity: 94.0%, specificity: 94.0%), respectively. Kandel et al. found that corneal aberration increases with KC severity, and indicators like corneal asymmetry index (SIb) and coma effectively detect subclinical KC (AUC > 0.90) [40]. Although the change of corneal thickness contributes to diagnosis (AUC = 0.77–0.94), its sensitivity and specificity are lower than those of aberration parameters (AUC = 0.71–0.95) [41]. Vertical coma consistently exhibited the greatest magnitude of HOA across all irregular cornea types [42]. In this study, the epithelial HOAs RMS and coma demonstrated excellent diagnostic accuracy, achieving AUC values exceeding 0.989, with very high sensitivity (95.9%–97.3%) and specificity (94.5%–96.7%). These parameters also effectively diagnosed the mild stage of KC, maintaining high AUC values of 0.976 and 0.974, with high sensitivity of 92.9% and 95.2%, specificity of 92.3% and 96.7%, respectively. For KC, most stromal parameters showed AUC values between 0.70 and 0.90, which were generally lower than those observed for epithelial aberrations. Given that KC may manifest in different corneal regions among individuals, epithelial or stromal thickness alone may be insufficient for reliable diagnosis. In contrast, epithelial wavefront analysis provides a comprehensive assessment of the corneal epithelium and offers broader applicability across various KC presentations. As a result, the epithelial HOAs RMS and coma could be independent indicators for discriminating KC eyes from healthy corneas. Elkitkat et al. demonstrated that BCV indices differed significantly between healthy individuals and those with KC using MS-39 (P < 0.001) [43]. The analysis yielded an AUC value of 0.994, along with a sensitivity of 97.7% and a specificity of 98.4%. Moreover, the EAI-KC discriminant function constructed by the epithelial HOAs RMS, coma and SA, attained the highest AUC value of 0.996 with 98.6% sensitivity and 98.9% specificity for diagnosing KC. Only the EAI-KC (AUC = 0.996) exceeds BCV of 0.994, indicating exceptional KC diagnostic performance. Coma (AUC = 0.990) and HOAs RMS (AUC = 0.989) are slightly lower, while total RMS, trefoil, SA, and Ast II fall below the value. The preference for "Epithelial Wavefront Analysis" arises from its ability to provide a comprehensive and sensitive assessment of the corneal epithelium, which is particularly valuable in the early detection of KC. The epithelium undergoes compensatory remodeling—thinning over areas of stromal protrusion and thickening in surrounding regions. This adaptive behavior allows epithelial aberrations to highlight subtle, region-specific abnormalities that may be missed by stromal thickness maps, especially given the variable presentation of KC across different individuals. Furthermore, epithelial wavefront analysis offers a user-friendly diagnostic approach, enhancing its clinical utility for detecting early or subclinical cases.

Our study is not without limitations. The MS-39 device has been shown to produce reproducible measurements in both healthy and keratoconic eyes [44,45,46,47]. However, since the repeatability of epithelial aberrations has not yet been investigated, we used mean values from three consecutive measurements to obtain reliable data [48]. Moreover, the EAI discriminant function was not cross-validated with an independent dataset due to the small sample size, limiting its generalizability. Future studies should also investigate the longitudinal changes of these indices and their potential to predict disease progression and exploring the combination of epithelial aberration analysis with other technologies.