Artificial Intelligence-Driven Detection of LASIK Using Corneal Optical Coherence Tomography Maps

Area of Science:

• Ophthalmology
• Medical Imaging
• Artificial Intelligence

Background:

• Identifying prior laser-assisted in situ keratomileusis (LASIK) is crucial for accurate ophthalmic assessments.
• Corneal optical coherence tomography (OCT) provides detailed cross-sectional images of the cornea.
• Distinguishing between healthy and post-LASIK corneas is essential for patient management.

Purpose of the Study:

• To develop and validate a convolutional neural network (CNN) for detecting LASIK surgery history.
• To utilize corneal OCT maps as input for the AI model.
• To assess the model's performance in identifying both myopic and hyperopic LASIK.

Main Methods:

• A lightweight CNN was trained using five types of corneal OCT maps (pachymetry, epithelial thickness, posterior mean curvature, anterior axial power, anterior stroma reflectance).
• Data included OCT scans from healthy individuals and patients with prior myopic or hyperopic LASIK.
• The model was evaluated using fivefold cross-validation and an independent out-of-sample test set.

Main Results:

• The CNN achieved high accuracy in cross-validation, with balanced accuracy of 90.2% for myopic LASIK and 90.2% for hyperopic LASIK.
• Sensitivity reached 93.5% for myopic LASIK and 90.2% for hyperopic LASIK.
• Area under the ROC curve (AUC) exceeded 97.8% for myopic and 98.2% for hyperopic LASIK detection. All eyes were correctly classified in the out-of-sample test.

Conclusions:

• A lightweight CNN model utilizing corneal OCT maps is effective for detecting LASIK history.
• This AI-powered approach offers a valuable tool for ophthalmic diagnostics.
• Artificial intelligence-assisted OCT imaging can improve the management of patients with prior LASIK undergoing procedures like cataract surgery.