Deep iterative vessel segmentation in OCT angiography.

Area of Science:

• Ophthalmology
• Medical Imaging
• Computational Biology

Background:

• Accurate retinal vessel segmentation is crucial for planning regenerative therapies in vitreo-retinal surgery.
• Optical coherence tomography angiography (OCT-A) provides high-detail visualization of macular vasculature, essential for surgical targeting.
• Existing segmentation methods may lack the precision required for advanced surgical applications.

Purpose of the Study:

• To develop and validate a novel convolutional neural network approach for precise retinal vessel segmentation in OCT-A images.
• To delineate the Superficial Vascular Plexus (SVP) in 2D Maximum Intensity Projections (MIP) of OCT-A.
• To assess the performance of the proposed method against existing techniques and evaluate its generalization capabilities.

Main Methods:

• Utilized convolutional neural networks (CNNs) for iterative refinement of vessel segmentation in OCT-A.
• Applied the method to 2D MIP reconstructions of OCT-A data.
• Conducted extensive experimental analysis on data from 50 subjects (surgical and healthy).

Main Results:

• The proposed CNN approach demonstrated superior performance compared to baseline networks and graph-based methods.
• Achieved favorable results in segmenting the Superficial Vascular Plexus.
• Demonstrated successful generalization to 3D OCT-A segmentation and narrower fields of view.

Conclusions:

• The developed method offers high-precision retinal vessel segmentation from OCT-A, suitable for surgical planning.
• The findings support the use of this technique for detailed mapping of macular vasculature.
• Future applications include leveraging vessel maps for intraoperative navigation in vitreo-retinal surgery.