Deep Learning-Based Estimation of Implantable Collamer Lens Vault Using Optical Coherence Tomography.

Area of Science:

• Ophthalmology
• Artificial Intelligence
• Medical Imaging

Background:

• Accurate measurement of implantable collamer lens (ICL) vault is crucial for assessing outcomes after refractive surgery.
• Manual vault measurement using anterior segment optical coherence tomography (AS-OCT) can be time-consuming and subject to inter-observer variability.

Purpose of the Study:

• To develop and validate a deep learning neural network for automated ICL vault measurement.
• To assess the accuracy and reliability of the AI model compared to manual measurements.

Main Methods:

• A deep learning network was trained using transfer learning on 2647 AS-OCT scans from 139 eyes.
• The model was validated on a separate test set of 191 scans.
• Performance was evaluated using metrics including Mean Absolute Percentage Error (MAPE), Mean Absolute Error (MAE), Root Mean Squared Error (RMSE), and Pearson correlation coefficient (r).

Main Results:

• The model achieved high accuracy with an MAPE of 3.42%, MAE of 15.82 µm, and RMSE of 18.85 µm.
• A strong positive correlation was observed between the AI-estimated vault and manual measurements (r = +0.98, R² = +0.96).
• No significant difference was found between AI-estimated and technician-measured vaults (P = .064).

Conclusions:

• A deep learning neural network can accurately and reliably compute ICL vault from AS-OCT scans.
• The AI model overcomes limitations of imbalanced and limited training data.
• This algorithm can serve as a valuable tool for postoperative assessment in ICL surgery.