Artificial intelligence-assisted projection-resolved optical coherence tomographic angiography (aiPR-OCTA).

Area of Science:

• Ophthalmology
• Medical Imaging
• Artificial Intelligence

Background:

• Optical coherence tomographic angiography (OCTA) is crucial for visualizing retinal vasculature.
• Projection-resolved OCTA (PR-OCTA) enhances image quality but can retain artifacts.
• Artifacts and signal loss limit the diagnostic accuracy of current PR-OCTA methods.

Purpose of the Study:

• To develop and evaluate an artificial intelligence-based PR-OCTA (aiPR-OCTA) algorithm.
• To improve artifact removal and flow signal preservation in PR-OCTA.
• To maintain anatomical details at the capillary level in retinal imaging.

Main Methods:

• A convolutional neural network was trained to generate projection-resolved OCTA volumes.
• Structural OCT and OCTA data served as inputs for the AI model.
• A dataset of 126 normal eyes was used for algorithm evaluation.

Main Results:

• The aiPR-OCTA algorithm demonstrated superior artifact removal compared to rule-based methods.
• Enhanced preservation of in situ flow signals and capillary-level anatomical details was observed.
• The aiPR-OCTA method achieved a higher flow signal-to-noise ratio (fSNR) and reduced background artifacts.

Conclusions:

• AI-driven PR-OCTA significantly improves image quality in retinal angiography.
• aiPR-OCTA offers a promising tool for enhanced diagnosis and research in ophthalmology.
• This advanced imaging technique preserves critical vascular information for clinical applications.