Ectasia Risk Model: A Novel Method Without Cut-off Point Based on Artificial Intelligence Improves Detection of...

Area of Science:

• Ophthalmology
• Artificial Intelligence
• Machine Learning

Background:

• Corneal ectasia is a serious complication after laser in situ keratomileusis (LASIK).
• Accurate risk assessment is crucial for patient selection and preventing ectasia, especially in cases with normal preoperative topography.
• Current models often rely on individual risk factor cut-offs, which may limit their predictive accuracy.

Purpose of the Study:

• To develop and validate a novel artificial intelligence (AI) based ectasia risk model.
• To create an integrated risk assessment method that does not rely on individual risk factor cut-off points.
• To improve the differentiation of patients at higher risk of developing ectasia, particularly those with normal preoperative topography.

Main Methods:

• A comparative case-control study involving 339 eyes (65 ectasia, 274 control) with normal preoperative topography.
• An AI model was developed using 20 features, including engineered variables derived from known risk factors.
• The model assessed risk based on the interaction of variables, not isolated cut-off values, with t-distributed stochastic neighbor embedding (t-SNE) used for visualization.

Main Results:

• The final AI model selected two original variables (percent tissue altered, corneal thickness) and two engineered variables (derivative percent tissue altered, age-weighted value).
• The AI-based model demonstrated a significantly greater ability (P < .0001) to differentiate between high-risk and low-risk patients compared to methods using isolated variables.
• t-SNE visualization confirmed the superior discriminative power of the AI model in separating ectatic and control groups.

Conclusions:

• A novel AI-based model effectively integrates multiple risk factors for ectasia without relying on cut-off points.
• This integrated approach enhances the accurate identification of patients at higher risk of ectasia.
• The AI model shows promise for improving preoperative risk stratification in LASIK surgery.