Clinical Utility of Artificial Intelligence Algorithms to Enhance Wide-Field Optical Coherence Tomography Angiography...

Area of Science:

• Ophthalmology
• Medical Imaging
• Artificial Intelligence

Background:

• Optical Coherence Tomography Angiography (OCT-A) is crucial for retinal imaging.
• Standard OCT-A images can be affected by noise, potentially impacting clinical assessment.
• Deep learning algorithms offer potential solutions for image noise reduction.

Purpose of the Study:

• To evaluate the clinical utility of deep learning denoising algorithms applied to wide-field OCT-A images.
• To compare the diagnostic quality of original versus denoised OCT-A scans for retinal vascular layers.

Main Methods:

• Retrospective case-series of 49 patients with wide-field OCT-A macula scans.
• Comparison of original and deep learning denoised images by retinal experts.
• Assessment focused on Superficial Capillary Plexus (SCP) and Deep Capillary Plexus (DCP).

Main Results:

• Denoised images were preferred for clinical assessment in 98% of Superficial Capillary Plexus evaluations.
• Denoised images were preferred in 35% of Deep Capillary Plexus evaluations.
• No new artifacts or loss of anatomical detail were observed in denoised images.

Conclusions:

• Deep learning denoising enhances the clinical utility of wide-field OCT-A images, especially for SCP assessment.
• The denoising feature of the Canon Xephilio OCT-A1 provides clinically preferable scans.
• The algorithm effectively reduces noise without compromising image integrity.