Automatic segmentation of anterior segment optical coherence tomography images

Area of Science:

• Ophthalmology
• Biomedical Engineering
• Medical Imaging

Background:

• Accurate measurement of the anterior segment geometry is crucial for ophthalmic applications.
• Optical coherence tomography (OCT) provides quantitative data but requires precise segmentation.
• Existing segmentation methods for OCT images may lack accuracy and reliability.

Purpose of the Study:

• To develop and validate a novel automated technique for segmenting the anterior and posterior corneal boundaries in OCT images.
• To improve the accuracy and reliability of quantitative measurements of the anterior segment geometry.
• To assess the performance of the new technique against established methods.

Main Methods:

• A three-step technique involving artifact removal, cornea localization via thresholding, and a level set-based shape prior segmentation model.
• Segmentation of anterior and posterior corneal boundaries using the proposed model.
• Comparison of the technique's performance against previous methods using manual annotations on 33 normal eye OCT images.

Main Results:

• The new technique demonstrated significantly improved segmentation concordance compared to previous methods.
• A mean Dice's similarity coefficient greater than 0.92 was achieved, indicating high accuracy.
• The method successfully segmented corneal boundaries in optical coherence tomography images.

Conclusions:

• The developed level set-based shape prior segmentation model offers accurate and reliable segmentation of corneal boundaries in OCT images.
• This technique has the potential to enhance quantitative measurements of anterior segment geometry.
• Accurate geometric data is vital for applications like numerical simulations of ocular biomechanics.