A two-stage framework for optical coherence tomography angiography image quality improvement.

Area of Science:

• Ophthalmology
• Medical Imaging
• Biomedical Engineering

Background:

• Optical Coherence Tomography Angiography (OCTA) is crucial for visualizing retinal and conjunctival microvasculature.
• Acquired OCTA images, especially Anterior Segment OCTA (AS-OCTA), often suffer from poor quality due to artifacts and low contrast.
• These image quality issues hinder accurate interpretation and diagnosis.

Purpose of the Study:

• To develop and evaluate a two-stage framework for enhancing OCTA image quality.
• To address stripe artifacts and low contrast in OCTA and AS-OCTA images.
• To improve the accuracy of subsequent vascular segmentation and clinical diagnosis.

Main Methods:

• A two-stage framework was proposed, including a de-striping stage and a re-enhancing stage.
• A Stripe Removal Net (SR-Net) with a novel de-striping objective function was used to suppress stripe noise.
• A Perceptual Structure Generative Adversarial Network (PS-GAN) with cyclic perceptual and structure loss was employed for image re-enhancement.

Main Results:

• The proposed framework demonstrated promising enhancement performance on synthetic and real AS-OCTA datasets.
• The de-striping and re-enhancing stages effectively improved image quality by reducing artifacts and enhancing vessel structures.
• Enhanced OCTA images led to improved performance for both conventional and deep learning-based vessel segmentation methods.

Conclusions:

• The proposed two-stage framework significantly enhances OCTA and AS-OCTA image quality.
• Improved image quality facilitates more accurate microvasculature analysis and diagnostic interpretation.
• This method holds potential for advancing clinical applications of OCTA imaging.