DERMA-OCTA: A Comprehensive Dataset and Preprocessing Pipeline for Dermatological OCTA Vessel Segmentation.

Area of Science:

• Dermatology
• Medical Imaging
• Computer Vision

Background:

• Optical coherence tomography angiography (OCTA) is a key non-invasive technique for skin vascular imaging.
• Standardized analysis methods and annotated datasets are lacking for dermatological OCTA.
• Automated analysis tools require extensive, high-quality annotated data for development.

Purpose of the Study:

• To introduce DERMA-OCTA, the first open-access dataset for dermatological OCTA.
• To provide a comprehensive resource for training deep learning models and benchmarking algorithms.
• To facilitate research in non-invasive skin vascular imaging.

Main Methods:

• Collected 330 volumetric OCTA scans from 74 subjects with diverse skin conditions.
• Included original 2D and 3D OCTA data, processed versions (5 methods), and reference segmentations.
• Generated 2D and 3D segmentation labels using U-Net architecture.

Main Results:

• Established the DERMA-OCTA dataset, featuring high-resolution annotated OCTA data.
• Dataset includes raw and processed scans with detailed segmentation labels for various skin pathologies.
• The U-Net architecture was employed for generating both 2D and 3D segmentation masks.

Conclusions:

• DERMA-OCTA is a valuable, freely downloadable resource for advancing dermatological OCTA research.
• The dataset supports the development of deep learning models for automated skin vascular analysis.
• Facilitates benchmarking of segmentation algorithms and promotes non-invasive skin imaging research.