Skin capillary extraction technique based on independent component analysis and Frangi filter using videomicroscopy

Area of Science:

• Dermatology
• Biomedical Engineering
• Medical Imaging

Background:

• Videomicroscopy offers a non-invasive method for observing skin capillaries.
• Existing methods require enhancement for accurate skin region analysis.
• Blood vessel extraction filtering combined with color component separation is a promising pre-processing approach.

Purpose of the Study:

• To develop and validate an algorithm for extracting skin capillary regions.
• To quantitatively analyze physiological responses of skin capillaries.
• To assess the algorithm's accuracy using ground truth comparisons.

Main Methods:

• Acquisition of skin images via videomicroscopy.
• Development of an algorithm combining independent component analysis (ICA) and the Frangi filter.
• Extraction and comparison of capillary blood vessel images against ground truth data.
• Evaluation of physiological responses to stimuli like CO2 water, local heating, and methyl nicotinate.

Main Results:

• The algorithm achieved a Dice coefficient of 0.82 when compared to ground truth images, indicating high accuracy.
• Exposure to carbon dioxide (CO2) water for 2 minutes significantly increased the capillary area.
• The study demonstrated the algorithm's ability to detect changes in capillary area under different physiological conditions.

Conclusions:

• A novel algorithm utilizing ICA and the Frangi filter effectively extracts capillaries from skin images.
• The proposed algorithm enables quantitative analysis of physiological changes in superficial skin capillaries.
• This technique holds potential for non-invasive monitoring of microvascular function.