Automated Epidermis Segmentation in Ultrasound Skin Images
The automated system for epidermis segmentation in ultrasound images of skin is described in this paper. The method consists of two main parts: US probe membrane segmentation and epidermis segmentation. The fuzzy C-means clustering is employed at the initial stage leading to probe membrane segmentation using fuzzy connectedness technique. Then, the upper (external) epidermis boundary is detected and adjusted using connectivity and line variability analysis. The lower (internal) boundary is obtained by shifting the upper edge by a constant vertical width determined adaptively during the experiments. The method is evaluated using a dataset of 13 US images of two registration depths. The validation relies on a ground truth delineations of the epidermis provided by two independent experts. The mean Hausdorff distances of 0.118 mm and 0.145 mm were obtained for the external and internal epidermis boundaries, respectively, with the mean Dice index for the epidermis region at 0.848.
KeywordsSkin imaging Skin layers High-resolution ultrasound Image segmentation
This research was supported partially by the Polish National Science Centre (Narodowe Centrum Nauki) grant No. UMO-2016/21/B/ST7/02236 and partially by the Polish Ministry of Science and Silesian University of Technology statutory financial support No. BK-209/RIB1/2018.
- 3.Ravichandra, G., Arjun, S., Ajmal, S., Manjunath, S., Ayshath, S.: High resolution ultrasonography in dermatology: a psoriasis experience. Indian J. Basic Appl. Med. Res. 5(2), 121–125 (2016)Google Scholar
- 8.Bugdol, M., Czajkowska, J., Pietka, E.: A novel model-based approach to left ventricle segmentation. In: 2012 Computing in Cardiology, pp. 561–564, September 2012Google Scholar
- 10.Badura, P., Kawa, J., Czajkowska, J., Rudzki, M., Pietka, E.: Fuzzy Connectedness in segmentation of medical images. A look at the pros and cons. In: International Conference on Fuzzy Computation Theory and Applications (FCTA 2011), pp. 486–492, October 2011Google Scholar