Convolutional Neural Network-based Optical Coherence Tomography (OCT) A-scan Segmentation and Tracking Platform using...

Area of Science:

• Ophthalmic imaging
• Medical artificial intelligence
• Computational modeling

Background:

• Accurate segmentation of corneal structures in Optical Coherence Tomography (OCT) images is crucial for diagnosing and monitoring ocular diseases.
• Current segmentation methods may face limitations in accuracy and efficiency, necessitating advanced computational approaches.

Purpose of the Study:

• To develop and validate a parallel Monte Carlo simulation platform for generating synthetic OCT cornea images.
• To train a convolutional neural network (CNN) using these synthetic images for automated cornea segmentation.
• To evaluate the performance of the trained CNN on ex-vivo cornea A-scan images.

Main Methods:

• Implementation of a parallel Monte Carlo simulation to generate realistic OCT cornea images.
• Development and training of a convolutional neural network (CNN) for image segmentation.
• Validation of the CNN's segmentation accuracy on ex-vivo cornea A-scan datasets.

Main Results:

• The Monte Carlo simulation platform successfully generated high-fidelity synthetic OCT cornea images.
• The trained CNN demonstrated improved accuracy in segmenting corneal layers from ex-vivo A-scan images compared to existing methods.
• The simulation-based training approach proved effective for enhancing AI model performance in ophthalmic imaging.

Conclusions:

• The developed parallel Monte Carlo simulation platform is a valuable tool for generating training data for AI in ophthalmology.
• AI-powered segmentation using simulation-generated data shows significant potential for improving the analysis of OCT cornea images.
• This approach facilitates the advancement of automated diagnostic tools in eye care.