To log in, click the teal "Login" button in the upper right-hand corner of this page. If you are logged in but still do not have access, please check your 2026 Annual Meeting registration.
Moderator
Scott M. MacRae, MD, ABO
Panelists
Stephen D. Klyce, PhD; Helen K. Wu, MD, ABO
Viewing Papers
Expand a paper title to the right to view the paper abstract and authors. Use the video link to jump to that poster in the session.
Presenting Author
Ceren Ece Semiz, MD
Co-Authors
Faruk Semiz (MD), Fetih Furkan Arslan (MD), Njomza H (MD), Merjem Purelku (MD, PhD), Gamze Tanrivedi (MD, PhD)
Purpose
To investigate the presence and potential of telocytes in the stroma extracted via SMILE surgery from keratoconus patients several years post-treatment.
Methods
This retrospective study included 60 eyes. Inclusion criteria were stage 2-3 keratoconus (Amsler-Krumeich), corneal thickness 300-419 µm, and indication for penetrating keratoplasty; exclusions included active anterior segment pathology, corneal thickness ?420 µm, or prior corneal/anterior segment surgery.Fresh convex-shaped lenticules, obtained from myopic SMILE donors, were transplanted into recipient corneas. Six lenticule samples (one untreated keratoconic eye and five treated eyes, 1-5 years post-surgery) were examined using CD34 immunohistochemistry and double immunofluorescence (CD34/PDGFR-?, CD34/c-kit) to identify telocytes and progenitor cells.
Results
In untreated keratoconic lenticule, CD34 staining revealed irregular, thinned collagen and reduced stromal cell density with degenerative changes. Post-surgery, cell density (including CD34-positive progenitors and telocytes) increased at 1-3 years but declined by 4-5 years, as confirmed by double immunofluorescence. At 5 years, UDVA improved from 1.34 to 0.58 logMAR, CDVA from 1.11 to 0.43 logMAR, CCT from 367.5 to 480.4 µm, while K1/K2 decreased from 53.16/59.93 D to 51.35/54.89 D, and Q-value from ?0.75 to ?0.47 (all p < 0.001).
Conclusion
Fresh lenticule transplantation is a safe, cost-effective, and reliable procedure that enhances corneal thickness, improves visual acuity, and facilitates corneal repair. This regenerative process is driven by active progenitor cells and telocytes, which play a crucial role in renewing stromal tissue.
Presenting Author
GÖzdenur Erba?, MD
Co-Authors
Ferah Özçelik (MD), Cigdem Altan (MD), Nilay kandemir Besek (MD)
Purpose
This study aims to investigate the correlation between visual acuity achieved with rigid contact lenses and pinhole test results in patients with keratoconus, and to examine how these outcomes are associated with cone localization and cone morphology.
Methods
A retrospective analysis was performed on the records of 35 keratoconus patients (16 females, 19 males) treated from 2021 to 2024 with clear corneas, no hydrops scarring, and fitted with rigid contact lenses, including a total of fifty eyes. Uncorrected visual acuity (UCVA), best corrected visual acuity with spectacles (BCVA), best corrected visual acuity with rigid contact lenses (BCCLVA), and pinhole-corrected visual acuity (PHCVA) were recorded(Snellen). Cone localization (central/paracentral) and cone type (nipple/oval/global) were classified.
Results
The mean values for UCVA, BCVA, BCCLVA, and PHCVA were 0.09 ± 0.10, 0.30 ± 0.18, 0.66 ± 0.19, and 0.42 ± 0.19, respectively (p < 0.0001). Moderate-to-strong positive correlations were found between BCVA and PHCVA (r = 0.577, p < 0.001), and between PHCVA and BCCLVA (r = 0.446, p = 0.001). A significant correlation between PHCVA and BCCLVA was observed in central cones (r = 0.409, p = 0.006), but not in paracentral cones. BCCLVA improvements compared to PHCVA averaged 0.25 ± 0.21 and varied by cone type, with the greatest improvement in nipple cones (p = 0.002).
Conclusion
The pinhole test is a practical, cost-effective predictor of visual improvement in keratoconus, especially for central cones. Visual improvement with rigid contact lenses compared to the pinhole test varied by cone morphology, with greatest gain in nipple cones. Findings highlight individualized lens fitting importance.
Presenting Author
Gaurav Prakash, MD, FRCS
Co-Authors
Marah Shelbayh (MD), Vishal Jhanji (MD)
Purpose
To evaluate how popular tomographic indices perform across increasing corneal thinning and ectasia severity in normal and keratoconus (KC) eyes, and to assess potential bias in keratoconus detection.
Methods
This retrospective study included 478 KC and 1330 normal eyes (1 eye/patient). Thirteen tomographic indices from the Pentacam® (Oculus), including Belin-Ambrosio D (BAD-D) score, its subcomponents, and central keratoconus index, were analyzed. Indices were regressed against the thinnest corneal thickness (TCT) and maximum keratometry (Kmax) to model increasing severity of thinning and steepening.
Results
In normal eyes, BAD-D showed a 2.5-fold stronger predictive fit with TCT (R�=0.23, p<0.01) than with Kmax (R�=0.09, p<0.01). No significant correlation existed between TCT and Kmax in normal eyes. In KC eyes, predictive fits were stronger and more balanced: BAD-D with TCT (R�=0.54, p<0.01) and with Kmax (R�=0.63, p<0.01). Similar patterns were observed for other non-BAD indices.
Conclusion
In normal eyes, BAD-D scores can be higher disproportionately in thinner corneas. This can lead to potential misclassification as a Keratoconus suspect.
Presenting Author
Fares H Alrefai, BSc
Co-Authors
Karanpreet Multani (MD), Kamran Riaz (MD)
Purpose
To evaluate the impact of socioeconomic factors, demographic factors, and medical comorbidities on severity of keratoconus (KCN) at presentation. To identify predictors of progression and surgical intervention, and to assess continuity of care in a large, diverse academic cohort.
Methods
A retrospective chart review of 21,157 encounters (2014-2021) at a tertiary academic center identified 797 patients with KCN (BAD-D ?1.6, Kmax ?47D). Manual review confirmed diagnosis and assessed severity (K2 ?52D), progression (?1D K2 increase), surgical interventions, and loss to follow-up over two years. Demographic, socioeconomic (insurance, education, life expectancy, language), and clinical factors were collected. Logistic regression and generalized linear mixed models were used, with covariates selected from univariate analysis and final models by backward elimination.
Results
Of 797 patients with KCN (60.1% male; mean age 41.7 ± 16.5 years), 57.5% self-identified as racial/ethnic minorities. Severe disease at presentation was common (79.7%) and was associated with younger age (P < 0.001), minority status (P = 0.032), lower education (P = 0.05), lower life expectancy (P = 0.011), non-commercial insurance (P = 0.0075), and English as a second language (ESL; P = 0.034). Younger age was associated with a higher likelihood of progression (P = 0.029) and undergoing surgery within two years (P=0.004); ESL was also linked to surgery (P = 0.013). The two-year loss to follow-up rate was 65.1%, significantly higher than that reported in published community cohorts.
Conclusion
Minority status, socioeconomic disadvantage, and language barriers were linked to advanced KCN at presentation and higher surgical needs. Younger age and ESL predicted progression. High attrition underscores systemic barriers. Early detection and targeted follow-up are crucial in reducing disparities.
Presenting Author
Ramadhan Ahmed, BSc
Co-Authors
Nakul Shekhawat (MD, MPH), Mahad Mohamed (BSc), Kamini Reddy (PharmD)
Purpose
To assess the agreement and repeatability of corneal tomographic measurements in keratoconus eyes using the Anterion Cornea App compared to Pentacam HR.
Methods
Eyes with keratoconus and no other significant corneal pathology underwent imaging with swept-source OCT (Anterion Cornea App, Heidelberg Engineering, Germany) and Scheimpflug tomography (Pentacam HR, Oculus, Germany). Each eye was imaged three times per device by the same technician. Measured parameters included anterior/posterior keratometry (K1, K2, Kavg, ?Ks, Kmax), central corneal thickness (CCT), and thinnest pachymetry (TP). Repeatability was assessed using intraclass correlation coefficient (ICC) and coefficient of variation (CV). Agreement was evaluated using Pearson correlation (?) and generalized estimating equations.
Results
Data collection is ongoing. Thus far, 26 eyes from 14 patients have been imaged with both devices. Anterion showed ICC >0.950 & CV <10% for all values. Pentacam HR had ICC >0.920 & CV <10% except for posterior ?K (ICC 0.733, CV 23.0%) & anterior ?K (CV 21.5%). Agreement between devices was near perfect (? >0.980) for all values except anterior ?K (?=0.871) & posterior ?K (?=0.645). No significant differences were found in most mean values, except for anterior K1, anterior ?K, CCT, & TP. Compared to Pentacam, Anterion measured greater anterior ?K (3.94 vs 2.98, p=0.004) & lower anterior K1 (47.66 vs 48.51, p=0.009), CCT (472.87 vs 479.28, p=0.002), & TP (450.77 vs 463.88, p<0.001).
Conclusion
Data collected thus far shows that in keratoconus eyes, Anterion has excellent reliability for all tomographic values while Pentacam HR has less reliability for anterior & posterior ?K. Both devices have excellent agreement with no significant differences for most values, though Anterion measures steeper anterior ?K & lower anterior K1, CCT, & TP.
Presenting Author
mehmet Bugra Tumtas, MD
Co-Authors
Sibel Ahmet (FEBO, FRCSEd, MD), Nilay kandemir Besek (MD), Ahmet Kirgiz (FEBO), Furkan Kirik (MD), Gülay Yalç?nkaya çak?r (FEBO)
Purpose
To identify predictors of progression in pediatric keratoconus cases followed for 10 years after accelerated corneal cross-linking (ACXL).
Methods
This retrospective cohort included 95 eyes from 61 patients (mean follow-up 112±25 months). Eyes were classified as progressed (n:16) versus not progressed (n:79) at 10 years. Baseline demographics, visual acuities, spherical-equivalent refraction (SER), and corneal topography (thinnest corneal thickness [TCT], K1, K2, Kmax, Kast [K1-K2]) were compared Mann-Whitney U and Fisher's exact tests, as appropriate. Univariate logistic regression screened predictors, and a prespecified multivariable model included age, SER, TCT, and Kast. Regression standard errors were cluster-robust at the patient level. Model discrimination was quantified by the receiver operating characteristic (ROC) analysis.
Results
Groups were similar at baseline for gender, visual acuities, K1, K2, and Kmax, and Amsler-Krumeich stage (p>0.05). The progression group was younger (p=0.029) and had thinner TCT (p=0.013). In the univariate logistic regression models, age alone predicted lower odds of progression (OR=0.71; p=0.042). In the multivariable model (adjusted for age, SER, TCT per 50 µm, and Kast), older age (OR=0.64; p=0.009), higher Kast (OR=0.66; p=0.002), and thicker TCT (OR=0.37 per 50 µm; p=0.047) were independently associated with reduced odds of 10-year progression, whereas baseline SER was not significant (p=0.176). The model demonstrated good discrimination (Area Under the Curve (AUC)=0.816).
Conclusion
At 10-year follow-up after ACXL in pediatric keratoconus, younger baseline age, thinner cornea, and lower keratometric astigmatism were associated with a higher risk of progression. Preoperative age, TCT, and Kast should be carefully considered in the long-term follow-up of pediatric keratoconus patients after ACXL.
Presenting Author
Michael B. Raizman, MD
Co-Authors
Tomas Navratil (PhD), Eric Brown (BSc), David Usher (PhD, MSc), Jane Ellen Giamporcaro (MS, BSc)
Purpose
A significant number of keratoconus patients go undiagnosed and untreated due to eyecare professionals' lack of access to topography/tomography equipment or unfamiliarity with the keratoconus' signs/symptoms or diagnostic methods. This presentation outlines the design and development of a novel, portable device developed to screen for keratoconus.
Methods
The keratoconus screening device is a handheld device to be used by eyecare professionals to screen for keratoconus. This device illuminates the cornea with Placido rings, captures the anterior image, and applies real-time algorithm to the captured image to determine the probability of keratoconus. A series of studies have been developed to validate the device's sensitivity and specificity. Study #1 (feasibility) will evaluate data collection, alignment and correlation parameters. Study #2 (formative) will evaluate the ability of the classifier. Study #3 will evaluate the sensitivity and specificity to validate the classifier.
Results
Findings from Study #1 involving 2 investigators and 140 eyes (112 healthy eyes and 28 diagnosed with keratoconus) showed a 99.3% (139/140) rate of successful automated image capture, with a median of 3.9 seconds to collect a valid eye scan. For rating ease of performing guided handheld alignment via a 5-point scale of 1 (“very dissatisfied”) to 5 (“very satisfied”), the majority of users were satisfied and the average rating was 4.1. Study #2 has been initiated and Study #3 is planned for March.
Conclusion
Based on results from Study #1, the keratoconus screening device demonstrates automated image capture with effective alignment, offering a streamlined, user-friendly solution to aid eyecare professionals in identifying patients who are at high risk for keratoconus.
Presenting Author
Alexandra Gil, BSc
Co-Authors
Leslie Vargas (MD), Alexander Gasper (BA), Daniela Spies (BA), Rosali Nodarse (BSc), William Trattler (MD)
Purpose
To quantify the risk of keratoconus progression among patients delaying corneal cross-linking (CXL) for different time intervals: 6-12 months, 12-18 months, and >18 months.
Methods
A retrospective study chart review was performed on 363 eyes diagnosed with keratoconus that delayed CXL treatment for at least 6 months. Pentacam topography was reviewed, and KMax values from initial diagnosis were compared with the pre-procedure measurements. Progression was defined as an increase of ?1.0 diopter (D) in KMax. Patients were stratified into 3 groups based on delay duration: short (6-12 months), medium (12-18 months), and long (>18 months).
Results
In the 6-12 month delay group, (N=185) 30.8% of eyes progressed, with a mean ?KMax= 0.95D. In the 12-18 month group, (N=71) 38.0% of eyes progressed, mean ?KMax= 1.22D. In the >18 month group, (N=107) 49.5% of eyes progressed, mean ?KMax= 2.20D. Although longer delays were associated with greater risk of progression and larger mean steepening in all groups, the >18 month group had significantly more progression compared with the 6-12 month group (p= 0.002). When considering only eyes that progressed, the mean ?KMax was 3.56 D in the 6-12 month group, 3.41 D in the 12-18 month group, and 4.87 D in the >18 month group.
Conclusion
Keratoconus progression occurred across all delay intervals, with greatest risk and steepening when CXL was delayed >18 months. Outcomes at 12-18 months resembled 6-12 months, but waiting longer substantially increased risk. Minimizing treatment delays is crucial to reduce progression.
Presenting Author
Jaron Sanchez, DO
Co-Authors
Mohammad Soleimani (MD)
Purpose
To investigate healthcare disparities associated with keratoconus (KCN) patients receiving treatment (corneal cross-linking [CXL] and keratoplasty), as well as comorbidities associated with KCN.
Methods
A total of 3224 patients from the University of Illinois Hospital & Health Sciences System (UI-Health) database from 2020 to 2024 were examined, including 1612 patients with KCN, and 1612 ophthalmology patients as a control group. Multivariable and univariable logistic regression were performed to evaluate the associations between sociodemographic traits and rates of CXL and keratoplasty. Best corrected visual acuity (BCVA) and manifest cylinder were used as indicators of disease impact. Comorbid disease rates were compared to a 1:1 distance-matched control group. Main outcome measures were odds ratio of undergoing keratoplasty and CXL, and prevalence of comorbid conditions.
Results
Female individuals received less keratoplasty than male individuals (odds ratio [OR]=0.55, P<0.001). Black individuals received less CXL than White individuals (OR=0.68, P<0.05), as did individuals with Medicaid (OR=0.27, P<0.0001) or no insurance (OR=0.41, P<0.001) compared to those with commercial insurance. Individuals from socially vulnerable neighborhoods received less CXL (OR=0.56, P<0.01) and keratoplasty (OR=0.66, P<0.05). Black female individuals were the most vulnerable, undergoing fewer procedures than White female (OR=0.58, P<0.01) and Black male (OR=0.65, P<0.05) individuals. Black and Hispanic/Latin-X individuals presented with more severe disease (P<0.01, P<0.0001).
Conclusion
Significant sociodemographic disparities exist in the treatment of KCN. Although further research is necessary, addressing these disparities is crucial for ensuring equitable access to care.
Presenting Author
Salomon Merikansky, MD
Co-Authors
Guillermo Amescua (MD), Heather Durkee (PhD), Juan Navia (MD), Jaime Martinez Martinez (MD), Noel Ziebarth (PhD), James Lai (MD)
Purpose
Riboflavin CXL is limited to corneas >400 ?m due to endothelial risk, and modified protocols for thinner corneas still carry complications. Rose bengal (RB) penetrates ~100 ?m and may be safer. This study used AFM to assess biomechanical stiffness changes after RB-CXL
Methods
21 pairs of ex vivo human corneas (<10 days post-mortem) were collected and de-epithelialized per the Dresden protocol. In each pair, the OS served as untreated control, while the OD was soaked in 0.1% RB for 30 minutes and irradiated with green light (5.4 J/cm�) at either 6 mW/cm� for 15 min (n=10) or 17 mW/cm� for 5 min (n=11). Young's modulus was measured by AFM at Bowman's layer and 50 ?m depth, the latter obtained using a 50 ?m microkeratome cut. OCT was performed before and after each cut to confirm corneal thickness.
Results
At the 50 ?m depth, the Young's modulus of elasticity of corneas in the 17 mW/cm2 irradiation group (25.7 ± 16.7 kPa) were significantly higher than the contralateral control corneas (10.4 ± 5.0 kPa), p<0.001. No differences were observed between the 17 mW/cm2 and 6 mW/cm2 groups or 6 mW/cm2 group and the control at 50 ?m, p?0.05. There were no differences in Young's modulus between any of the groups at Bowman's layer, which was expected.
Conclusion
RB-CXL with high-intensity green light (17 mW/cm�) significantly increased corneal stiffness at 50 ?m depth, while no differences were seen at Bowman's layer between 6 and 17 mW/cm�. These results suggest RB-CXL preferentially strengthens deeper corneal layers, offering promise for treating thin corneas at reduced endothelial risk.
Presenting Author
Andrew Mihalache, None
Co-Authors
Ryan Huang (MD, MSc), Marko Popovic (MD), Margarita Safir (MD), Adi Einan-Lifshitz (MD), Clara Chan (MD, FRCSC)
Purpose
This population-based study aimed to investigate the relationship between attention deficit hyperactivity disorder (ADHD) and keratoconus (KC) in a large, diverse adult cohort in the United States.
Methods
Using data from the National Institutes of Health All of Us Research Program, individuals diagnosed with ADHD were propensity score-matched (1:3) to unique controls without ADHD based on age, body mass index (BMI), sex, sociodemographic characteristics, and smoking status. Univariable and multivariable conditional logistic regression models evaluated the association between ADHD and KC, with adjustments for psychiatric comorbidities, atopic conditions, obstructive sleep apnea, ADHD medications, conjunctivitis, and tear film-related disorders. An interaction term analysis assessed potential effect modification by sex. Odds ratios (ORs) with 95% confidence intervals (CIs) were reported.
Results
40,360 matched participants were included, comprising 10,090 individuals with ADHD and 30,270 without ADHD (mean age: 42.3 ± 15.6 years; mean BMI: 30.5 ± 8.4 kg/m2; 63.1% female). KC prevalence was significantly higher among individuals with ADHD (33/10,090, 0.33%) compared to matched controls (42/30,270, 0.14%), yielding an unadjusted odds ratio of 2.36 (95% CI=[1.49, 3.72], p<0.01). Participant sex did not modify this association (p=0.38). After multivariable adjustment, the association between ADHD and KC remained significant (OR=2.43, 95% CI=[1.28, 4.61], p=0.01), with tear film-related disorders emerging as an independent predictor of KC (OR=5.37, 95% CI=[2.62-11.01], p<0.01).
Conclusion
In this large, diverse adult cohort in the United States, individuals with ADHD were more than twice as likely to have KC, independent of multiple confounders. Tear film-related disorders, which may represent a proxy for eye rubbing, also emerged as a predictor. These findings underscore the value of KC risk stratification in individuals with ADHD.
Presenting Author
Andrew Mihalache, None
Co-Authors
Ryan Huang (MD, MSc), Prem Nichani (MD, MSc, BSc), Marko Popovic (MD), Clara Chan (MD, FRCSC)
Purpose
This study evaluated the performance of the latest version of ChatGPT-5 in detecting keratoconus (KC) from corneal topography maps.
Methods
In September 2025, ChatGPT-5 (Instant mode) was prompted with an open-source dataset of corneal topography maps from Scheimpflug imaging systems (Pentacam®), graded by three specialists as KC, KC suspect, or normal. Each eye contributed seven maps: anterior/posterior eccentricity, anterior/posterior elevation, anterior/posterior sagittal curvature, and central corneal thickness. Outcomes included sensitivity, specificity, accuracy, positive predictive value (PPV), and negative predictive value (NPV). The primary analysis evaluated ChatGPT-5's ability to distinguish KC from normal eyes. Secondary analyses assessed (i) KC or KC suspect versus normal and (ii) KC versus KC suspect or normal.
Results
A total of 200 KC, 173 KC suspect, and 200 normal eyes were analyzed. In the primary analysis (KC versus normal), ChatGPT-5 achieved a sensitivity of 96.5% (95%CI=92.9%-98.6%), specificity of 77.0% (95%CI=70.5%-82.6%), accuracy of 86.8% (95%CI=83.0%-89.9%), PPV of 80.8% (95%CI=76.5%-84.4%), and NPV of 95.7% (95%CI=91.4%-97.9%). In the secondary analysis, when KC and KC suspects were grouped versus normal eyes, sensitivity and specificity declined to 74.3% (95%CI=69.5%-78.6%) and 75.5% (95%CI=68.9%-81.3%), respectively. In contrast, distinguishing KC from either KC suspect or normal eyes yielded higher sensitivity (93.0%, 95%CI=88.5%-96.1%) but lower specificity (62.5%, 95%CI=57.3%-67.4%).
Conclusion
ChatGPT-5 showed strong performance for rapidly distinguishing KC from normal eyes on corneal topography; however, performance declined when KC suspects were included, suggesting greater reliability in detecting definitive rather than borderline cases. Future research should validate the chatbot's ability to screen for KC under expert oversight.
Presenting Author
Bassel Hammoud, MD, MS
Co-Authors
Hongyuan Zhang (PhD), Bianca Susanna (MD), Barbara Dutra (MD, PhD), Giuliano Scarcelli (PhD), James Randleman (MD)
Purpose
To determine the proximity between the thinnest corneal point (TCP) and focal corneal weakening in normal, subclinical keratoconus (SKC), and manifest keratoconus (KC) eyes using motion-tracking Brillouin microscopy.
Methods
Prospective cross-sectional study. 95 eyes from 95 patients were evaluated: 40 from bilaterally normal patients (controls), 40 from patients with SKC, and 15 from patients with manifest KC. All patients underwent Scheimpflug tomography, anterior segment OCT (AS-OCT), and custom-built motion-tracking Brillouin (MTB) imaging. Mean and minimum MTB shift values were calculated within the anterior plateau region. Euclidean distances between the TCP (identified by AS-OCT) and the minimum Brillouin shift (MTB-Min) were determined. Receiver operating characteristic curves were generated for both metrics to determine the area under the curve (AUC), sensitivity, and specificity.
Results
No significant differences were found between groups for age, sex, KMean, or KMax. Subclinical keratoconus and KC eyes were significantly thinner than controls. Motion-tracking Brillouin minimum values were significantly lower in SKC and KC eyes compared with controls. Average distances between MTB-Min and TCP were 0.31 ± 0.16 mm (0.04-0.69 mm) in controls, 0.53 ± 0.28 mm (0.11-1.19 mm) in SKC, and 0.54 ± 0.35 mm (0.10-1.36 mm) for KC (P < 0.001). Motion-tracking Brillouin minimum values were within 1 mm of the TCP in 100% of control eyes, 92.5% of SKC eyes, and 86.7% of KC eyes (P = 0.1). MTB-min values fully differentiated SKC from controls (AUROC = 1).
Conclusion
Focal corneal weakening occurs in close proximity to the thinnest corneal point in SKC. This study provides the first experimental confirmation of this relationship and demonstrates focal mechanical localization in subclinical and early KC.
Presenting Author
Bassel Hammoud, MD, MS
Co-Authors
Jad Assaf (MD), Cynthia Roberts (PhD), Shady Awwad (MD), Zahi Wehbi (MD)
Purpose
To enhance the CLMI.X index for detecting keratoconus suspect (KCS) cases by incorporating advanced machine learning (ML) algorithms.
Methods
This study included 352 eyes divided into 3 categories: normal (NL) (n = 133), KCS (n = 77), and keratoconus (KC) (n = 142). Imaging was obtained using the Galilei dual Scheimpflug-Placido system. Eleven variables, identical to those used in the original CLMI.X, were employed to train and test multiple ML models, including logistic regression (LR) and more advanced algorithms, for classification of corneas into NL, KCS, and KC. SHapley Additive exPlanations (SHAP) analysis was performed to identify influential variables for differentiating KCS from NL eyes, both for the full KCS group and a subgroup of topographically and tomographically borderline cases.
Results
The original CLMI.X demonstrated high sensitivity (100%) and specificity (99%) for detecting KC but low sensitivity (4%) for KCS. Training the CLMI.X-AI using LR on 3 classes improved KCS sensitivity to 68%. Incorporating advanced artificial intelligence algorithms further increased KCS sensitivity to 75%, with an overall accuracy of 89%. SHapley Additive exPlanations analysis identified the posterior axial, posterior instantaneous, and anterior axial curvature maps as the most influential variables, alongside zonal pachymetry and posterior elevation. For topographically and tomographically NL fellow eyes, SHAP analysis revealed significant influence for zonal pachymetry.
Conclusion
The CLMI.X-AI demonstrates substantially improved performance in detecting KCS compared with the original CLMI.X. Its ML enhanced decision-making and 3-class training make it a more clinically relevant and robust tool for early KC detection.
Presenting Author
Sunayana Narasimha Murthy, MBBS, MS
Co-Authors
Rohit Shetty (FRCS), Abha Shah (MS)
Purpose
To evaluate clinical outcomes of ray-tracing-guided customized ablation combined with accelerated cross-linking in progressive keratoconus
Methods
A Prospective interventional study where 106 eyes with progressive keratoconus underwent customized ablation using ray-tracing software with high-frequency excimer laser (2 kHz), followed by accelerated corneal cross-linking. Assessments at 1, 3, and 6 months included visual acuity, corneal tomography, and higher-order aberration(HOA) analysis.
Results
At 1 month, Kmax reduced by 6.3 ± 3.9 D (p < 0.05) and Kmean by 3.2 ± 1.5 D (p < 0.05), stable through 6 months. UCDVA and BCDVA improved ?3 lines (p < 0.05). Mean spherical equivalent decreased by 1.4 D (p < 0.05). HOA reduced from 2.32 µm to 1.43 µm (p < 0.05).
Conclusion
Ray-tracing-guided ablation with cross-linking provides effective corneal regularization and visual rehabilitation. This approach offers tissue-conserving keratoconus treatment.
Presenting Author
Barbara Dutra, MD, PhD
Co-Authors
Bassel Hammoud (MD, MS), Bianca Susanna (MD), Hongyuan Zhang (PhD), Giuliano Scarcelli (PhD), James Randleman (MD)
Purpose
To determine the biomechanical efficacy of standard and modified corneal cross-linking (CXL) protocols. We hypothesize that Brillouin microscopy enables quantitative, noninvasive assessment of spatial and temporal stiffening in vivo, providing a direct measure of CXL impact.
Methods
This is a prospective, randomized animal study in rabbits allocated to sham control or CXL with different irradiation parameters, including accelerated and transepithelial protocols. Brillouin microscopy was performed preoperatively and longitudinally up to six months, along with tomography and intraocular pressure. Primary outcomes were changes in Brillouin-derived modulus over time. Analyses compared treated groups to controls and across protocols using regression, adjusting for corneal thickness.
Results
TBD
Conclusion
TBD
