Deep-learning visualization enhancement method for optical coherence tomography angiography in dermatology.

Area of Science:

• Dermatology
• Medical Imaging
• Artificial Intelligence

Background:

• Dermatological OCTA imaging quality is often poor due to skin's scattering properties and limited acquisition.
• Previous deep learning applications for OCTA image enhancement have been limited in dermatology.
• High-quality ground truth data for training deep learning models is difficult to obtain.

Purpose of the Study:

• To develop a robust deep learning method for enhancing skin OCTA images.
• To generate suitable datasets for training deep learning models for dermatological OCTA.
• To improve the visualization of cutaneous vasculature in OCTA.

Main Methods:

• A swept-source skin OCTA system was used to acquire low- and high-quality images.
• A generative adversarial network (GAN) model named vascular visualization enhancement GAN was proposed.
• Optimized data augmentation and perceptual content loss were employed for training.

Main Results:

• The proposed GAN model significantly enhanced the quality of skin OCTA images.
• Quantitative and qualitative comparisons demonstrated the superiority of the developed method.
• Effective image enhancement was achieved even with limited training data.

Conclusions:

• The developed deep learning method effectively enhances dermatological OCTA images.
• This approach addresses the challenges of low image quality in skin OCTA.
• The method shows promise for improving diagnostic capabilities in dermatological imaging.