Screening of prior refractive surgery by a wavelet-based neural network

Area of Science:

• Ophthalmology
• Biomedical Engineering
• Artificial Intelligence

Background:

• Corneal topography is essential for diagnosing and managing refractive errors.
• Screening for previous refractive surgery is crucial for accurate eye assessments.
• Objective methods are needed to reliably identify post-refractive surgery corneas.

Purpose of the Study:

• To develop and validate an objective screening method for previous myopic refractive surgery.
• To utilize corneal topography data for automated detection of refractive surgery.
• To assess the accuracy of a wavelet-based neural network for this purpose.

Main Methods:

• Videokeratography data from normal and post-refractive surgery corneas were analyzed.
• Axial curvature values were extracted to form 1-dimensional waveforms.
• Multiresolution wavelet decomposition and a backpropagation neural network were employed for classification.

Main Results:

• The system achieved 99.3% overall accuracy in detecting previous myopic refractive surgery.
• Sensitivity was 99.1%, and specificity was 100% for identifying treated corneas.
• The neural network effectively distinguished between normal and post-refractive surgery eyes.

Conclusions:

• A 1-dimensional wavelet-based neural network is an effective and accurate tool for screening previous myopic refractive surgery.
• This objective method reliably differentiates post-refractive surgery eyes from normal corneas.
• The system's performance highlights the potential of AI in ophthalmic diagnostics.