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
Marcony R. Santhiago, MD, PhD
Panelists
Winston D Chamberlain, MD, PhD, ABO; Olivia L. Lee, 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
Lourdes Haydaw, BSc
Purpose
A retrospective study to determine the incidence of posterior ocular astigmatism (POA) after removal of topography guided excimer laser ablation utilizing topography measured astigmatism and axis to remove anterior corneal astigmatism and measure residual astigmatism in 1500 eyes.
Methods
WaveLight Contoura was performed with topography measured astigmatism and axis, and all inaccurate outcomes were analyzed for the presence of low topographic astigmatism and significant residual manifest astigmatism. A secondary enhancement procedure was then performed, and measurements taken to measure the presence of topography measured astigmatism and the lack of manifest astigmatism to confirm that POA had been present.
Results
Out of the 1500 eyes, 44 were confirmed to have POA, resulting in an incidence of 2.93%. The average amount of manifest astigmatism in these eyes post-primary procedure was 1.13D, and the average topography measured astigmatism was 0.39D. After secondary enhancement with wavefront optimized ablation, the average amount of post-operative astigmatism was 0.01D, and the average amount of topography measured astigmatism was 1.11D. Post-secondary enhancement procedure, average higher-order aberration was reduced 9%.
Conclusion
The incidence of 2.93% is lower than prior estimated amounts of posterior ocular astigmatism.
Presenting Author
James B Randleman, MD
Purpose
To compare the efficacy of Motion-tracking (MT) Brillouin microscopy and multimodal corneal imaging for keratoconus suspect identification.
Methods
Prospective cross-sectional study evaluating 56 eyes from 56 patients, including 28 eyes from 28 bilaterally normal control patients and 28 eyes from 28 keratoconus suspect patients. All patients underwent Scheimpflug tomography, AS-OCT), Corvis ST, and MT Brillouin microscopy using a custom-built device. All readily available metrics from each clinical device were evaluated. Motion-tracking (MT) Brillouin microscopy data included mean and focal minimum MT Brillouin 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 differences were found between groups in any demographic parameter except thinnest corneal thickness (556 mm vs. 526 mm; P < 0.001). MT Brillouin outperformed all other imaging technologies in group differentiation. The best performing metrics evaluated included MT Brillouin minimum (AUC = 0.999, Sensitivity 100%, Specificity 96%) and mean (AUC 0.981, Sensitivity 93%, Specificity 100%), Corvis TBI (AUC 0.906, Sensitivity 71%, Specificity 96%), PRFI (AUC 0.870, Sensitivity 64%, Specificity 96%), KI (AUC 0.873, Sensitivity 61%, Specificity 96%), and IHD (AUC 0.859, Sensitivity 64%, Specificity 89%). The best epithelial metric, Epi 5mm ST-IN, performed poorly (AUC = 0.641).
Conclusion
Motion-tracking Brillouin microscopy metrics effectively identified focal corneal weakening in keratoconus suspects, outperformed all multimodal imaging metrics, and performed optimally at differentiating keratoconus suspect eyes from normal controls.
Presenting Author
Luis Andino F. Santos, MD, MBA
Co-Authors
Robert Edward Ang (MD), Emerson Cruz (MD, MPH)
Purpose
To compare the LASIK eligibility assessments of GPT-4o from Pentacam data against two refractive surgeons, and to characterize discrimination, calibration, agreement, and directional biases.
Methods
Retrospective, single-center, observational per-eye study of patients screened from June-December 2024. Pentacam BAD-D and four-map displays were de-identified and assessed by GPT-4o using a standardized prompt, using three stochastic runs per eye with majority aggregation. Each surgeon independently provided binary eligibility and confidence. Primary metrics: sensitivity/specificity, PPV/NPV, AUC, Brier score, Cohen's ?, Fleiss' ? (overall), McNemar's test for paired directional differences. Model repeatability was summarized with ICC (3,1) (descriptive for binary outputs).
Results
453 eyes (233 patients, mean age 33.07 ± 10.82) were included. Surgeon eligibility: Surgeon 1 (S1): 85.9%, Surgeon 2 (S2): 53.6%; GPT-4o: 45.7%. Versus S1: sensitivity 53% (95% CI 48-58), specificity 97% (89-100), AUC 0.748 (0.692-0.803), Brier 0.107; McNemar = 176.1 (P < .001), indicating a conservative tilt. Versus S2: sensitivity 60% (54-66), specificity 71% (64-77), AUC 0.655 (0.606-0.705), Brier 0.225; McNemar = 7.76 (P = .005), indicating a smaller conservative shift. Agreement was fair (Fleiss'?: 0.23, P < .001. Cohen's ?: GPT vs S1 = 0.23, P < .001; GPT vs S2 = 0.31, P < .001; inter-surgeon = 0.29, P < .001). Repeatability across three GPT runs was high (ICC=0.84, 95% CI 0.81-0.87).
Conclusion
GPT-4o showed moderate discrimination, good probabilistic accuracy, and a safety-leaning profile (higher specificity, lower sensitivity), supporting use as a conservative adjunctive second reader rather than a replacement for clinician judgment.
Presenting Author
Betul Dumlu, MD
Co-Authors
Sevgi Tongal (MD, FEBO), Dilan Colak (FEBO), ?akir Ar?man (FEBO), Ahmet Kirgiz (FEBO)
Purpose
To compare photopic and mesopic chord ? values between keratoconus (KC) patients and healthy individuals, assess their correlation with biometric and aberrometric parameters, and evaluate their diagnostic value in KC detection and preoperative planning for refractive surgery.
Methods
In this cross-sectional study, 51 eyes with keratoconus and 50 healthy eyes were examined using the Topcon Aladdin Biometer and Corneal Topographer HW3.0. Chord ? values were calculated manually under both photopic and mesopic lighting conditions according to the Euclidean formula, based on pupil center and corneal light reflex coordinates. We compared photopic and mesopic chord ? values and their associations with biometric and aberrometric parameters. Statistical analyses were performed using SPSS software version 26.0(IBM, NY, USA), including Mann-Whitney U test, Spearman correlation, ROC curve analysis, and multivariate linear regression to assess diagnostic thresholds and parameters.
Results
Both photopic and mesopic chord ? values were significantly higher in the keratoconus group compared to controls (photopic: 0.309 ± 0.10 mm vs. 0.247 ± 0.08 mm, p < 0.001; mesopic: 0.272 ± 0.10 mm vs. 0.226 ± 0.09 mm, p = 0.003). ROC analysis revealed moderate diagnostic capacity: Photopic chord ? > 0.315 mm; AUC = 0.656, OR = 5.91 Mesopic chord ? > 0.275 mm; AUC = 0.615, OR = 3.33 In healthy eyes, chord ? correlated negatively with anterior chamber depth and axial length, suggesting anatomical variance. In keratoconus, chord ? correlated positively with vertical pupil shift and negatively with coma aberration under mesopic conditions.
Conclusion
Chord ? is significantly increased in keratoconus and reflects both geometric misalignment and optical distortion.While it lacks strong standalone diagnostic power,it may serve as a supportive parameter in borderline KC cases and assist in preoperative refractive surgery planning,especially where pupil-optic axis misalignment is a concern.
Presenting Author
Hatem A Elsersawy, MRCSEd
Purpose
To improve diagnostic accuracy in keratoconus and refractive surgery planning using an AI-orchestrated, multi-Ai-agents workflow that integrates multimodal imaging and patient history. Task-specific agents with LLM coordination enhance risk stratification, diagnostic certainty, and individualized treatment planning.
Methods
A retrospective study of 50 keratoconus and refractive surgery cases was conducted using a three-agent, AI-orchestrated workflow. Agents included: (1) General Ophthalmology (patient history, risk factors), (2) Imaging (Pentacam, AS-OCT, epithelial mapping, PTA), and (3) Anterior Segment specialist (diagnosis and surgical planning). Multiple LLMs coordinated task execution. Outcomes were measured as diagnostic accuracy, risk stratification consistency, surgical recommendation reliability, and model variability, benchmarked against general-purpose AI models.
Results
General Ophthalmology Agent: Accuracy of patient history: 84% (LLM range: ±6%) Risk factor consistency: 78% (Some misunderstood steroid-induced IOP risk) Pentacam & AS-OCT Agent: Imaging accuracy: 87% (LLM variability: ±5%) Detection of keratoconus: 92% (Enhanced in structured models) PTA accuracy: 81% (Borderline values resulted in variation) Anterior Segment Specialist Agent: Agreement on diagnosis: 89% (LLM variability: ±7%) Surgery recommendation: 76% (Variability in borderline LASIK cases)Total: 85% accuracy, LLM variation: ±9%, best time: 11.2s, worst time: 22.4s. Results emphasize LLM variation, indicating varying models for different agents.
Conclusion
An orchestrated, multi-agent AI workflow enhances diagnostic accuracy and clinical safety in refractive surgery and keratoconus. Task-specific agents consistently outperformed general models for risk stratification and decision-making. Future steps include multicenter validation and EMR integration.
Presenting Author
Damien Gatinel, PhD, MD
Co-Authors
Jacques Malet (PhD)
Purpose
To assess the clinical value of converting ocular wavefront data (µm) into vergence maps in diopters (D) and decomposing these maps into dioptric basis functions whose coefficients are expressed in diopters, enabling cl
Methods
Aberrometry (slit scanning skiascopy) was obtained over a 6.0?mm pupil. Wavefronts were transformed into vergence (D) by radial differentiation and mapped across the pupil. Each map was decomposed into dioptric functions (e.g., defocus/astigmatism, coma, trefoil, spherical aberration), yielding per?mode coefficients in diopters and an RMS vergence (D) of the HOA burden. Objective sphere?cylinder?axis (S/C/A) was derived from the central dioptric field to reflect paraxial optics. Clinical utility was illustrated in three eyes: normal myopic, keratoconic, and post?myopic LASIK.
Results
In a normal eye, objective refraction matched manifest within ?0.5 D and HOA burden was low (RMS ?0.4 D). In keratoconus, dioptric maps revealed inferior steepening with ?15 D peak-to-valley, dominated by coma (+6 D) and spherical aberration (+2 D); SE error vs manifest was ?0.9 D and HOA RMS ?5-6 D, highlighting the limits of sphero-cylindrical correction. Post-myopic LASIK showed a centro-peripheral gradient consistent with negative spherical aberration; SE error was ?0.1 D with RMS ?0.8-0.9 D. Across cases, dioptric coefficients in D allowed clinicians to rank HOAs, anticipate visual symptoms, and prioritize treatment targets.
Conclusion
Expressing aberrometry in diopters provides coefficients and RMS values that directly quantify the optical weight of each aberration. This facilitates clinical interpretation, helps judge higher-order effects, supports objective refraction consistent with manifest, guides ablation or IOL planning, and improves patient communication.
Presenting Author
Stuti Chamola, MS
Co-Authors
Rohit Shetty (FRCS), Deepthi Honniganur (MS), Abhijith Roy (MS), Rahul Patil (MSc), Raghav Narasimhan (None)
Purpose
To compare polarization-sensitive OCT (PS-OCT) with Pentacam and MS-39 in classifying corneal conditions and to evaluate the performance of AI-based models in detecting early keratoconus.
Methods
In this cross-sectional, single-center study, 359 eyes were analyzed: 120 healthy, 109 subclinical keratoconus (SKC), and 130 keratoconus (KC). All eyes underwent imaging with PS-OCT, Pentacam, and MS-39. PS-OCT provided birefringence and sub-layer thickness profiles, while the other devices assessed corneal curvature, pachymetry, and aberrations. AI models were independently trained using data from each modality to classify eyes into diagnostic categories and assess reclassification of SKC cases.
Results
AI models demonstrated high diagnostic performance across modalities. For healthy eyes, accuracy was 85.8% (PS-OCT), 85.0% (Pentacam), and 84.2% (MS-39). For KC, accuracy was 100%, 99.2%, and 100%, respectively. Among 66 consistently diagnosed SKC eyes, PS-OCT reclassified 43 (39.5%) as healthy, a higher rate compared to Pentacam and MS-39. The PS-OCT AI model achieved an area under the curve (AUC) of 0.91 with 82% overall accuracy, indicating strong discriminatory ability in differentiating early ectatic changes.
Conclusion
PS-OCT performed comparably to conventional tomography and demonstrated superior ability in reclassifying early ectatic cases without reliance on corneal shape indices.AI-driven PS-OCT collagen mapping may enhance early keratoconus detection and provide an adjunctive tool for safer refractive surgery screening.
Presenting Author
Apeksha M. Kataria, DNB
Co-Authors
Rohit Shetty (FRCS), Arka subhra Ghosh (PhD), Nimisha Kumar (MSc, PhD)
Purpose
This study explores whether Lysyl Oxidase (LOX) expression can guide the need for accelerated corneal crosslinking after refractive surgery
Methods
A Prospective Interventional study including total of 210 eyes undergoing either lenticule surgery (n=106) or photorefractive keratectomy PRK (n=104) underwent corneal tomography and biomechanics scans using Pentacam, Corvis, and polarization-sensitive OCT (PS-OCT). The epithelium and lenticule extracted were analyzed within 90 minutes for LOXs and ColA1 expression using quantitative PCR. Eyes with low LOX expression received accelerated crosslinking (45mV) after one week. Patients were categorized into 3 groups based on LOX expression.
Results
65 eyes undergoing lenticule surgery (n=35) or PRK (n=30) had low LOX levels and underwent accelerated crosslinking. LOXs and ColA1 increased from group 1 to 3. In group 3, CBI and SP-A1 correlated significantly with LOX levels (p<0.05). At 3-year follow-up, all patients showed stable refraction and corneal biomechanics.
Conclusion
This study highlights LOX expression as a rapid diagnostic test for ectasia risk and a guide for accelerated crosslinking in refractive surgery.
Presenting Author
David Smadja, MD
Co-Authors
Itay Lavy (MD), Sara Sella (MD)
Purpose
To quantify the effective epithelial ablation rate in transepithelial PRK (transPRK), determine its impact on residual refractive error and develop a nomogram adjustment to improve visual and refractive predictability.
Methods
This was a two-phase study, with preliminary quantification of the epithelial ablation rate (EAR) in animal corneas using a standard transepithelial photoablation profile, from which we derived an EAR-based epithelial ablation factor (EAF). In a second step, consecutive myopic eyes were treated with transPRK using the Teneo317z laser (Bausch and Lomb). For each eye, central epithelial thickness was combined with the EAF to set the epithelial ablation depth and adjust stromal targets. Outcomes at 3, 6 and 12 months (UDVA, CDVA, MRSE, efficacy/safety indices, haze, enhancements) were analyzed with multivariable models (covariates: age, baseline MRSE, CCT, K, optical zone).
Results
In the animal phase, epithelial ablation rate (EAR) was 21% lower than the programmed value, leading to an EAR-based epithelial ablation factor (EAF) increasing central epithelial ablation by 21% in transPRK. In 63 eyes, at 12 months 95.3% were within ±0.50 D of target and 96.8% achieved 20/20 UDVA. No eyes were overcorrected and no eye lost ≥2 lines of CDVA. No significant haze or safety concerns were observed
Conclusion
TBD
Presenting Author
Pooja Khamar, MD, PhD
Co-Authors
Rohit Shetty (FRCS)
Purpose
To investigate how corneal biomechanics influence intraoperative opaque bubble layer (OBL) formation and lenticule dissection, and how these factors affect early visual recovery and quality of vision, offering insights into why some patients remain dissatisfied despite good visual acuity.
Methods
In this prospective study, 300 eyes underwent kerato-lenticule extraction (KLEX) by a single surgeon on the same platform. Preoperative workup included Pentacam AXL Wave for topography, Corvis ST for biomechanics (Young's modulus), and polarization-sensitive OCT for collagen and Bowman's thickness. Intraoperative OBL formation was graded (1-3), and lenticule dissection ease was scored (1-4). Postoperative outcomes were measured using the Quality of Vision (QoV) questionnaire at day 1, week 1, and month 1. Associations between biomechanical parameters, intraoperative grades, and visual recovery trajectories were analyzed.
Results
All eyes achieved 20/20 Snellen acuity on day 1. However, biomechanical differences influenced early satisfaction. In 110 eyes with higher elasticity (>100 N/m�) and thicker Bowman's membrane (>17 ?m), grade 2-3 OBL and difficult dissection (grade 3-4) were observed, correlating with delayed QoV recovery. In contrast, 190 eyes with lower elasticity (<90 N/m�), thinner Bowman's (<15 ?m), and normal collagen orientation showed grade 1 OBL, easy dissection (grade 1), and faster QoV recovery. Although both groups achieved comparable QoV by 1 month, early dissatisfaction strongly aligned with unfavorable biomechanical profiles.
Conclusion
Biomechanics, OBL formation, and dissection difficulty explain why some patients report slower recovery or dissatisfaction despite good acuity. Adapting laser energy, spot size, and track distance to biomechanical profiles may improve early QoV and enhance patient satisfaction.
Presenting Author
Nir Sorkin, MD
Co-Authors
Nir Gomel (MD)
Purpose
To establish the mathematic association between lens tilt, corneal astigmatism and manifest astigmatism for the purpose of predicting manifest astigmatism using values of corneal astigmatism and lens tilt.
Methods
A retrospective study including 600 eyes of 600 patients who underwent measurements using a swept-source OCT biometer (Eyestar 900, Haag Streit, Switzerland) that measures anterior keratometry, posterior keratometry and lens tilt. Manifest refraction was performed by a certified optometrist. Models were constructed to evaluate the association between lens tilt, anterior keratometry, posterior keratometry and manifest astigmatism. The model found was validated for its ability to predict manifest astigmatism using values of lens tilt, anterior keratometry and posterior keratometry.
Results
Mean lens tilt was 3.9±1.3°. The association between lens tilt and internal astigmatism (IA) was threshold-dependent. In eyes with tilt <5.0°, there was no significant correlation of tilt with IA (J0: r=0.01, p=0.74, J45: r=0.10, p=0.001), indicating tilt is not a primary predictor in this group. However, in eyes with tilt >5.0°, a graphic trend towards correlation emerged, growing stronger in eyes with tilt >6.0°, where J45 was borderline significant with increasing correlation coefficient (r=0.41, p=0.07).
Conclusion
Lens tilt was not a universal predictor of manifest astigmatism; in majority of eyes with <5.0° tilt, IA was not mainly driven by lenticular tilt. However, beyond a threshold of ~5.0°, tilt became a more significant determinant of IA, suggesting that surgical planning may need to account for tilt beyond this threshold.
Presenting Author
Marcony R. Santhiago, MD, PhD
Co-Authors
Claudia Morgado (MD), Nicole Larivoir (MD), Lycia Maria Sampaio (MD), Juliana Santos (MD)
Purpose
To investigate the spatial relationship between epithelial thickness distribution and corneal topographic inferior-superior (IS) asymmetry, and to identify zone-specific epithelial metrics that best correlate with varying degrees of asymmetry.
Methods
A total of 526 eyes were analyzed. IS values were obtained from topography and used to stratify eyes into four groups. Twelve OCT-derived epithelial thickness difference metrics were calculated across central, temporal, and nasal zones in the 2-5 mm, 2-7 mm, and 2-9 mm annuli. Correlation analyses were performed across the full IS spectrum and separately for eyes with IS ? 1.4 and >1.4 diopters.
Results
Epithelial differences within the 2-5 mm annulus showed the strongest correlations with IS values. In eyes with IS ? 1.4, epithelial correlations were weak and positive, whereas in eyes with IS > 1.4, correlations became strong and negative, particularly in the inferior temporal-superior nasal axis (r = -0.68, p < 0.0001). This directional switch suggests a threshold-dependent epithelial remodeling response. Mean epithelial values were similar across Groups A-C (IS ? 1.4), with sharp deviations only emerging in Group D (IS > 1.4).
Conclusion
The relationship between epithelial thickness and IS asymmetry is nonlinear, with an inflection at 1.4 D. Remodeling is minimal in low asymmetry but becomes pronounced beyond this threshold. Zone-specific metrics, especially in the 2-5 mm region, may aid epithelial-based screening strategies.
Presenting Author
Ivan Gabric, MD
Purpose
To evaluate the accuracy and potential clinical benefits of Foresight, a corneal simulation software developed by Schwind Eye Tech Solutions. This platform predicts the postoperative corneal shape and guides customized ablation strategies for complex refractive cases, particularly in keratoconus.
Methods
A total of 60 eyes—30 treatment-naïve (“virgin”) and 30 keratoconus with prior corneal crosslinking (CXL)—were analyzed. Corneal imaging and tomography were performed with the MS39 (CSO Italia). Foresight was used to plan ablations on the Schwind Amaris 1050 laser, simulating postoperative corneal geometry and estimating higher-order aberrations, especially coma and astigmatism.
Results
Postoperative outcomes closely matched the Foresight simulations, demonstrating high accuracy in predicting corneal shape and refractive changes. The software’s ability to model coma and residual astigmatism was particularly noteworthy. In keratoconus eyes with previous CXL, unique ablation profiles generated via Foresight improved visual quality beyond conventional treatment protocols.
Conclusion
Foresight, integrating advanced simulation with tailored ablation planning, marks progress in personalized vision correction for virgin and complex keratoconus eyes. Though larger studies are needed, it shows promise for safe, precise, individualized outcomes when paired with Schwind Amaris 1050 and MS39 corneal imaging.
