Discrimination of Skin Layers from B-Scan Ultrasound Images Using Feature Matching and Neuro-Fuzzy Classifier

The measurement of skin thickness is an important method in diagnosis, control of treatment and studying the effects of drugs and their dosage on skin diseases. The development of high frequency ultrasound imaging systems in recent years has made the measurement of skin layers thickness possible. In this paper, a new method of identifying skin layers and measuring their thickness using B-scan ultrasound images is presented. The B-scan images of skin have been obtained using a Dermascan unit model C (Cortex Technology, Denmark) with a 20 MHz probe. In this research, feature matching and neuro-fuzzy classifier have been used to determine the boundaries between the skin layers. Feature matching is necessary for constructing the probability matrix followed by the neuro-fuzzy classifier which is used to extract the boundaries of skin layers. After finding these boundaries, a thickness of different skin layers is calculated. The proposed method has been tested on 50 ultrasound images. The results obtained for the distinction of different skin layers and the calculated skin thickness were compared against the result obtained from two dermatologists who examined the cases and the thickness calculated by histological technique, respectively. By applying this algorithm, different skin layers are extracted with 86% accuracy. The results of this method were also confirmed by histological experiments with a precision of one hundredth of a millimeter. Discrete cosine transform (DCT) coefficients feature has showed that it has the best performance for the identification of skin layers.Discussion and The neuro-fuzzy classifier can be used as a reliable tool for processing ultrasound images of skin. In contrast to existing methods, the proposed method doesnt require initialization and it has good noise immunity. Therefore, this method can be applied to clinical routine to find the effects of drugs precisely.