Purpose
To compare the diagnostic efficacy of motion-tracking (MT) Brillouin microscopy with multimodal corneal imaging for keratoconus suspect (KCS) identification.
Design
Prospective, cross-sectional diagnostic evaluation.
Methods
There were 56 eyes from 56 patients evaluated, including 28 eyes from 28 bilaterally normal control patients and 28 eyes from 28 keratoconus suspect patients. All patients underwent MT Brillouin microscopy using a custom-built device, Scheimpflug tomography, anterior segment optical coherence tomography (AS-OCT), and Corneal Visualization Scheimpflug Technology (Corvis ST). MT Brillouin microscopy data included mean (MTB-mean) and focal minimum (MTB-min) values within the anterior plateau region. The sensitivity, specificity, and area under the receiver operating characteristic curves (AUROC) were calculated for variables from all modalities.
Results
No significant demographic differences were found between groups except thinnest corneal thickness (556 μm Control vs 526 μm KCS; P <.001). MT Brillouin metrics MTB-min (AUC = 0.999, Sensitivity 100%, Specificity 96%) and MTB-mean (AUC 0.981, Sensitivity 93%, Specificity 100%) outperformed all other multimodal imaging metrics in KCS identification and group differentiation. MTB-min significantly outperformed all multimodal metrics, including the Corvis ST Tomographic and Biomechanical Index (TBI) (AUC 0.906, P =.018), Pentacam Random Forest Index (PRFI) (AUC 0.870, P =.005), Keratoconus Index (KI) (AUC 0.873, P =.005), and Index of Height Decentration (IHD) (AUC 0.859, P =.005). The best epithelial metric, Epi 5 mm ST-IN, performed poorly in group differentiation (AUC = 0.641).
Conclusions
Motion-tracking Brillouin microscopy metrics effectively identified focal corneal weakening in keratoconus suspects and significantly outperformed all multimodal imaging metrics in differentiating keratoconus suspect eyes from normal controls.
Refractive surgical screening prioritizes identification of keratoconus at the earliest possible (subclinical) stage. , Modern laser vision correction has demonstrated excellent outcomes in appropriate candidates , and compares favorably with contact lens wear in safety and efficacy, cost, and patient preference. However, current screening protocols remain inherently conservative, as only morphologic surrogates for underlying mechanical decompensation are available clinically. Yet, postoperative ectasia continues to occur, , while many otherwise reasonable candidates are excluded from surgery. Thus, the need to accurately characterize the risk of ectasia susceptibility remains essential.
Multiple screening devices, metrics, and strategies for subclinical keratoconus detection have been developed, evolving from reflection-based anterior curvature analysis alone ,,, to multimodal corneal imaging utilizing a variety of technologies. ,, Multimodal imaging approaches typically encompass some combination of Scheimpflug imaging, anterior segment optical coherence tomography (AS-OCT) with epithelial mapping, biomechanical measurements, and artificial intelligence (AI). Despite this, these approaches have demonstrated suboptimal efficacy in discriminating clinically relevant ectasia risk features from normal anatomic variation at the earliest stages. ,,,,,,
Recent developments in theory, modeling, and experimental data support the premise that localized mechanical weakening precedes morphologic alteration. Until recently, however, this premise has proven challenging to confirm. Motion-tracking Brillouin (MTB) microscopy offers a compelling approach. , This technique provides three-dimensionally localized mechanical measurements. , Focal anterior stromal MTB metrics have effectively discriminated normal control eyes from eyes with subclinical or early keratoconus and have outperformed morphologic metrics in group differentiation. ,,, This technique thus offers novel mechnical information currently unavailable using multimodal corneal imaging.
Given the critical importance of accurate screening protocols and the persistent gap in clinical imaging evaluations, this study evaluated the relative efficacy of MT Brillouin microscopy and multimodal clinical imaging for keratoconus suspect identification.
METHODS
This study was performed as part of an ongoing prospective study conducted at the Cleveland Clinic Cole Eye Institute, registered at clinicaltrials.gov (NCT04598932), and approved by the Cleveland Clinic Institutional Review Board (IRB# 20-355). The study adhered to the Declaration of Helsinki, and all patients provided written informed consent for study inclusion. Patients recruited for this arm of the study were evaluated between September 2023 and June 2025.
The study included patients aged 18 to 55 categorized into one of two groups based on routine clinical examination and corneal imaging: (1) patients with bilaterally normal corneas (Controls); (2) patients identified as keratoconus suspects (KCS). Only one randomly selected eye was included per patient. All patients underwent a complete ocular examination that included best corrected distance visual acuity (CDVA), manifest refraction, and slit lamp biomicroscopy. Corneal imaging was performed using Scheimpflug tomography (Pentacam HR; Oculus) anterior segment optical coherence tomography (AS-OCT; Avanti RTVue XR version 2018.1.1.63; Optovue) and dynamic corneal deformation imaging (Corvis ST, Oculus). Motion-tracking Brillouin microscopy was performed using a custom-built device.
Group classifications utilized the same criteria as reported in our prior studies. ,, Patients included in the control group were consecutive patients who presented for laser vision correction evaluation, had bilaterally normal slit lamp, < 2 diopters of central/inferior asymmetric anterior corneal steepening on tangential curvature maps, and were determined to be good candidates for all laser vision correction (LVC) procedures. The KCS cohort was comprised of consecutive patients who presented for LVC evaluation and were classified as KCS in both eyes, but who did not meet criteria for manifest keratoconus in either eye. We chose the term KC suspect for this cohort specifically following past recommendations because neither eye had manifest keratoconus. ,, All KCS patients had 20/20 or better CDVA. Patients in this group all showed suspicious but not definitive findings on corneal imaging, including > 2 diopters of central/inferior asymmetric anterior corneal steepening on tangential curvature maps. Similar to prior studies, ,, as there are to date no specific morphologic metrics/criteria that accurately differentiate normal control eyes from keratoconus suspect/subclinical keratoconus eyes at the earliest stages, we did not rely on specific numeric values for group classification. Exclusion criteria included prior ocular surgery, inadequate corneal imaging, or corneal scarring that could impact imaging acquisition. Multimodal corneal imaging for all patients (controls and KCS eyes) is included as Supplementary Material.
As described in previous studies, ,, corneal mechanical profiles could be quantified by the Brillouin shifts measured by a custom-built motion-tracking Brillouin microscope. This non-contact and non-perturbative technique co-registered the Brillouin shift values with their positions to correct patient motion blur during a series of 5-second depth scans with a 15 μm step size at different lateral locations. As reported in prior work, the average of a linear-fitted plateau region is selected as the Brillouin value for that location. ,, Brillouin shifts within the anterior plateau region were averaged to represent the Brillouin shift at each measured location. The plateau region was identified as the region between the start of the corneal measurements and the cross-section of the 2 fitting lines (plateau and slope regions). Approximately 40 locations with no specified grid pattern within an 8 mm diameter circle were used to create a corneal Brillouin map, taking 20 mins or less per patient. This was the same methodology followed in each of our prior analyses ,,,,, ; the imaging process remained unchanged from prior studies for consistency. The examiner (HZ) was masked to the patient’s group classification and did not have access to clinical imaging, including pachymetry mapping, prior to data acquisition and analysis. Following MTB imaging processing, the thinnest corneal point (TCP) was determined from AS-OCT imaging, while focal corneal weakening was identified using MTB imaging. The minimum Brillouin shift value within 1 mm of the TCP was identified as the focal MTB-Min by a separate examiner (BH).
Scheimpflug metrics evaluated included keratometry and pachymetry values, anterior and posterior elevation, asymmetry indices, Belin/Ambrosio display total deviation, and Ambrosio relational thickness. AS-OCT Metrics evaluated included epithelial thickness measurements as provided by the device. Corvis metrics evaluated included the Corvis Biomechanical Index (CBI), the Tomographic and Biomechanical Index (TBIv2), SP A1, Pentacam Random Forest Index (PRFI), ARTh, Integrated Radius, and Stress Strain Index (SSIv2).
STATISTICAL METHODS
POWER ANALYSIS & SAMPLE SIZE DETERMINATION
To establish the appropriate sample size for our experimental groups, we performed a calculation to ensure that all metrics had sufficient power to distinguish normal from KCS subject (primary analysis). For this we used data reported by Ambrosio et al, which compared different Corvis ST metrics, in particular CBI data as it was determined to be the less sensitive of the two metrics, in order to power the study for both CBI and TBI. In our study the subject cohort is slightly different (KCS vs control instead of VAE-NT vs control), but for the purpose of power analysis we assumed they would provide similar values. In that study, the normal group had mean CBI = 0.06 with SD = 0.14, while the subclinical group (called very asymmetric ectasia with normal topography, VAE-NT) had mean CBI = 0.41 with SD = 0.40. Therefore, planning a study that used CBI as continuous response variable from normal vs KCS subjects, if the true difference between groups was 0.35, and assuming normal distribution and the larger SD of 0.4; we needed to evaluate 28 KCS subjects and 28 control subjects to be able to reject the null hypothesis that the population means of the experimental and control groups were equal with probability (power) 0.9. The Type I error probability associated with the test of this null hypothesis was 0.05.
DATA ANALYSIS
Statistical analysis was performed using SPSS version 29 (IBM, Armonk, New York, USA). Chi-squared test was used to compare categorical variables between different groups. Continuous variables were described as mean ± SD (minimum—maximum). Normality was assessed using Shapiro-Wilk test. Normally distributed variables were then compared using analysis of variance (ANOVA), whereas non-normally distributed variables were compared using Kruskal–Wallis test. Receiver operating characteristic (ROC) curves were generated to compare the ability of various MT Brillouin, Scheimpflug, AS-OCT, And Corvis ST metrics to distinguish between control and KCS eyes. Differences in the area under the ROC curves were assessed using the DeLong test for correlated ROC curves, with MTB Min as the reference parameter. To account for multiple comparisons, a Bonferroni correction was performed, and all P -values were adjusted accordingly, with a significance level of = 0.05 used for all statistical tests.
RESULTS
The study evaluated 56 eyes from 56 patients divided into two groups: controls ( N = 28) and KCS ( N = 28). An additional eight eyes from 8 patients (5 controls, 3 KCS) were excluded due to poor imaging quality (patient movement and/or insufficient lateral scan coverage available for analysis). Patient demographics are shown in Table 1 . There were no major demographic significant differences between groups except for thinnest pachymetry, with KCS eyes being thinner by 30 μm on average. There were no correlations between pachymetry and MTB-Min Brillouin shift values in either group ( Table 2 ).
Table 1
Patient Demographics
| Demographics | Controls ( n = 28) | KC Suspects ( n = 28) | P |
|---|---|---|---|
| Age (y) | 32.71 ± 6.15 (23.00-44.00) | 32.89 ± 6.77 (22.00-48.00) | .918 |
| Sex (% Males) | 43 | 64 | .108 |
| Sphere (D) | −3.26 ± 2.17 (−8.50 to 1.50) | −3.80 ± 1.91 (−7.75 to − 1.25) | .323 |
| Cylinder (D) | 0.55 ± 0.64 (0.00-2.25) | 0.99 ± 0.70 (0.00-2.75) | .01 |
| MRSE (D) | −2.98 ± 2.10 (−7.38 to 1.50) | −3.31 ± 1.93 (−7.75 to − 0.62) | .548 |
| K Mean (D) | 43.08 ± 1.43 (39.70-45.60) | 43.77 ± 1.54 (40.20-46.30) | .088 |
| K Max (D) | 44.16 ± 1.60 (41.64-47.68) | 44.87 ± 1.48 (40.64-47.60) | .089 |
| Thinnest pachymetry (μm) | 555.64 ± 28.19 (502.00-615.00) | 525.29 ± 28.05 (483.00-593.00) | <.001 |
D = diopters; K mean = mean keratometry value; K max = maximum keratometry value; KC Suspects = keratoconus suspects; MRSE = manifest refraction spherical equivalent.
Table 2
Correlation Between the Minimum Brillouin Shift Values and Corneal Thickness
Figure 1 shows representative composite images showing Scheimpflug, AS-OCT, Corvis, and MT Brillouin imaging for Control and KCS groups. Group metric comparisons for each metric are shown for Scheimpflug ( Table 3 ), AS-OCT ( Table 4 ), Corvis ( Table 5 ), and MT Brillouin microscopy ( Table 6 ) modalities. All Corvis and MT Brillouin microscopy metrics were significantly different between groups. Most Scheimpflug metrics were significantly different between groups, except for Maximum Posterior Elevation. No epithelial thickness metrics were significantly different between groups. Figure 2 shows Brillouin shift values by depth for a control and KCS patient, demonstrating clear differences between patients in the anterior plateau region but with overlapping mechanical values in the middle and posterior stroma.
Composite images showing Scheimpflug, AS-OCT, Corvis, and MT Brillouin imaging in representative control (A) and KC suspect (B) eyes. In Figure 1 A, the control patient exhibits an amorphous pattern in tangential and axial curvature. Corneal pachymetry and elevation maps are unremarkable. There is increased index of height decentration (IHD), with no other remarkable findings in asymmetry indices, epithelial mapping, or Corvis indices, and the MT Brillouin map shows a normal mechanical map without significant lateral deviation. In Figure 1B, the KC suspect patient exhibits 2.7D of asymmetric inferior steepening obliquely, with >1.5D asymmetry notable in axial curvature. There is asymmetric anterior and posterior elevation with increased index of height decentration (IHD). There are no remarkable findings in epithelial mapping or Corvis indices. The MT Brillouin map shows greater lateral deviation with focal weakening adjacent to the thinnest corneal point.
Table 3
Scheimpflug Variables
| Variables | Controls ( n = 28) | KC Suspects ( n = 28) | P |
|---|---|---|---|
| Anterior Tangential K Max (D) | 44.51 ± 1.75 (40.50-47.50) | 45.65 ± 1.96 (40.80-50.90) | .025 |
| Maximum Anterior Elevation ( m) | 4.12 ± 0.99 (3.00-6.00) | 5.79 ± 1.99 (4.00-11.00) | .016 |
| Maximum Posterior Elevation ( m) | 11.12 ± 4.04 (5.00-20.00) | 13.95 ± 9.26 (6.00-52.00) | .267 |
| KISA% | 8.56 ± 14.73 (0.37-74.54) | 17.89 ± 19.54 (0.77-78.11) | .033 |
| IS-value | 0.18 ± 0.43 (−0.79 to 0.95) | 0.88 ± 0.55 (−0.34 to 2.10) | <.001 |
| ISV | 15.46 ± 4.73 (9.00-31.00) | 18.50 ± 3.82 (14.00-26.00) | .005 |
| IVA | 0.10 ± 0.05 (0.03-0.22) | 0.16 ± 0.05 (0.08-0.28) | <.001 |
| KI | 1.02 ± 0.01 (0.99-1.04) | 1.04 ± 0.02 (1.01-1.08) | <.001 |
| IHA | 4.09 ± 2.76 (0.30-11.50) | 8.39 ± 5.15 (1.00-16.70) | <.001 |
| IHD | 0.01 ± 0.01 (0.00-0.02) | 0.02 ± 0.01 (0.01-0.03) | <.001 |
| ARTmax | 482.29 ± 79.04 (349.00-634.00) | 406.07 ± 97.42 (254.00-650.00) | .002 |
| BAD-D | 0.68 ± 0.54 (−0.43 to 1.74) | 1.35 ± 0.82 (−0.03 to 3.13) | <.001 |
Art Max = Ambrosio’s relational thickness; BAD- D = Belin Ambrosio total deviation; IHA = index of height asymmetry; IHD = index of height decentration; IS value = inferior/superior value; ISV = index of area variance; IVA = index of vertical asymmetry; KC = keratoconus; K Max = maximum keratometry value; KI = keratoconus index.
Table 4
AS-OCT Epithelial Thickness Variables
| Variables | Controls ( n = 28) | KC Suspects ( n = 28) | P |
|---|---|---|---|
| Epi 5 mm I | 56.63 ± 3.50 (49.13-68.59) | 55.94 ± 3.80 (48.67-61.12) | .755 |
| Epi 5 mm IT | 55.82 ± 3.63 (49.24-68.39) | 54.67 ± 3.95 (46.57-60.44) | .527 |
| Epi 5 mm SN-IT | −0.95 ± 1.99 (−6.22 to 2.78) | −0.40 ± 2.46 (−4.96 to 5.30) | .371 |
| Epi 5 mm S-I | −1.83 ± 2.08 (−6.02 to 2.12) | −2.42 ± 2.93 (−7.96 to 3.64) | .401 |
| Epi 5 mm ST-IN | −1.37 ± 1.61 (−5.73 to 0.99) | −1.87 ± 2.16 (−6.35 to 3.51) | .083 |
| Epi 5 mm Sup-Inf | −1.12 ± 1.26 (−3.47 to 1.62) | −1.22 ± 1.83 (−4.58 to 2.90) | .810 |
| Epi Overall Min | 50.80 ± 3.42 (44.13-58.63) | 49.71 ± 3.24 (44.33-54.85) | .244 |
| Epi Median | 54.22 ± 3.39 (46.19-64.67) | 52.78 ± 2.91 (46.25-55.50) | .735 |
| Epi Min-Median | −3.42 ± 2.49 (−8.61 to 0.60) | −3.07 ± 1.89 (−9.47 to − 0.39) | .576 |
| Epi Max | 60.14 ± 4.50 (52.49-75.94) | 59.49 ± 4.12 (52.86-71.67) | .646 |
| Epi Min-Max | −9.34 ± 3.45 (−18.38 to − 3.99) | −9.78 ± 3.60 (−18.30 to − 5.97) | .741 |
| Epi Std Dev | 1.91 ± 0.80 (0.88-4.12) | 2.10 ± 0.66 (1.41-4.00) | .135 |
I = inferior; KC = keratoconus; N = nasal; S = superior; T = temporal.
Table 5
Corvis Variables
| Variables | Controls ( n = 28) | KC Suspects ( n = 28) | P |
|---|---|---|---|
| SP A1 | 124.81 ± 16.15 (97.42-162.33) | 104.97 ± 15.07 (83.96-134.72) | <.001 |
| CBI | 0.17 ± 0.17 (0.01-0.88) | 0.42 ± 0.24 (0.03-0.88) | <.001 |
| TBI | 0.13 ± 0.13 (0.01-0.44) | 0.53 ± 0.28 (0.07-0.98) | <.001 |
| PRFI | 0.04 ± 0.04 (0.00-0.14) | 0.22 ± 0.21 (0.02-0.91) | <.001 |
| ARTh | 585.47 ± 93.13 (367.50-758.73) | 525.14 ± 110.29 (373.21-917.63) | .006 |
| Integrated radius | 7.93 ± 1.11 (5.43-10.42) | 8.86 ± 0.99 (6.70-10.71) | .002 |
| SSI | 1.00 ± 0.13 (0.76-1.41) | 0.92 ± 0.14 (0.71-1.21) | .035 |
Stay updated, free articles. Join our Telegram channel
Full access? Get Clinical Tree
