Multimodal Predictive Modeling for Visual Quality Recovery After Keratorefractive Lenticule Extraction.

Area of Science:

• Ophthalmology
• Medical Imaging
• Artificial Intelligence

Background:

• Keratorefractive lenticule extraction (KLEx) is a surgical procedure to correct myopia.
• Assessing and predicting postoperative visual quality is crucial for patient outcomes.
• Early identification of factors influencing visual recovery can optimize surgical planning.

Purpose of the Study:

• To identify factors affecting visual quality recovery post-KLEx.
• To develop a predictive model for postoperative visual quality using multimodal data.
• To integrate radiomics and clinical data for enhanced prediction accuracy.

Main Methods:

• Prospective study of 210 eyes from 105 myopic patients undergoing KLEx.
• Objective Scatter Index (OSI) used to assess visual quality pre- and post-surgery.
• U-net deep learning model for lenticule image segmentation and radiomics feature extraction.
• Integration of radiomics, clinical, topographical, and biomechanical data into machine learning models.

Main Results:

• 30.95% of eyes showed poor early visual recovery.
• Significant differences in spherical equivalent, corneal curvature, and cutting depth noted between groups.
• Machine learning models achieved high predictive accuracy for OSI (AUC up to 0.99).
• Specific radiomics features significantly correlated with postoperative visual quality.

Conclusions:

• A predictive framework integrating intraoperative imaging, biomechanics, and densitometry was established.
• Radiomics and clinical parameters can identify high-risk patients for KLEx.
• This approach enables personalized surgical planning to improve visual outcomes.