Ai-Driven Biomechanics-Integrated Lenticule Extraction (Bilex): Predicting Tissue Response And Optimizing Surgical Outcomes
Published 2025 - 43rd Congress of the ESCRS
Reference: FP26.12 | Type: Free paper | DOI: 10.82333/ea86-v750
Authors: Raquel Rodrigo Fernández 1 , Lucía Ibares Frías 1 , Blanca García-Valcárcel González 1 , Clara Leyton Swinburn* 1 , Víctor Antón Modrego 1 , Enma Yesenia Marín Concha 1 , Julio Rafael Ruiz Batres 1 , María Chamorro González-Cuevas 1
1Ophthalmology,Hospital General Universitario Gregorio Marañón,Madrid,Spain
Purpose
To develop an artificial intelligence (AI) model that predicts intraoperative tissue response and postoperative visual quality in kerato-lenticule extraction (KLEX) by analyzing preoperative imaging parameters. The study aims to optimize surgical planning by correlating biomechanics with lenticule dissection ease, opaque bubble layer (OBL) formation, and early visual recovery.
Setting
A single-center prospective study conducted in a clinical research setting.
Methods
A total of 300 eyes undergoing KLEX were analyzed using preoperative imaging from Pentacam HR (tomography), MS-39 (epithelial thickness, collagen structure), and Corvis ST (corneal biomechanics). An AI model was trained to classify eyes into "optimal" and "sub-optimal" groups based on day-1 postoperative visual quality (QoV scores). Key predictors were identified using feature importance analysis. Intraoperative OBL formation and lenticule dissection difficulty were recorded and correlated with AI-predicted outcomes. The model’s performance was assessed using area under curve (AUC), accuracy, sensitivity, specificity, and F1-score.
Results
The AI model achieved an AUC of 0.91, with 98% sensitivity and 85.6% specificity. It accurately predicted 94.9% of optimal and 72.4% of sub-optimal outcomes. Strong predictors included SpA1, corneal biomechanical index (CBI), and deformation amplitude (DA) ratio (2mm). Higher corneal stiffness was linked to increased OBL (grades 2-3), tougher dissection (grades 3-4), and delayed QoV recovery. Eyes with normal biomechanics showed minimal OBL (grade 1), easier dissection (grade 1), and faster visual recovery. By one month, both groups exhibited comparable QoV scores.
Conclusions
AI-driven analysis of preoperative biomechanics enables predictive modeling of intraoperative challenges and early visual outcomes in KLEX. This approach facilitates personalized surgical planning by optimizing laser energy settings, track spacing, and patient selection, ultimately improving postoperative recovery and patient satisfaction. Future AI refinements may further enhance real-time decision-making in refractive surgery.