Skin Cancer Evaluation using Texture and Form Analysis of Sonograms

Abstract

For the pre-operative assessment of skin tumors, high resolution ultrasound systems utilizing frequencies between 7,5 and 20 MHz are increasingly used [1–5] Currently, efforts are focusing on the processing of sonographic images to obtain an objective interpretation of sonographic findings [6–11]. In order to differentiate between various skin tumors, the subjective impression of the sonographer, which is determined by the individual experience, needs to be translated into reproducible texture and form analysis of the tumors. Traditionally, the sonographic assessment of tumors is based upon reproducible sonographic patterns. Sonographic images are the result of complex interaction (reflection, absorption, scattering) of the tissue with the transmitted signal. This is reflected by the sonographic B-image [12]. For the sonographic assessment, the lateral extension as well as that into deeper layers of the tumor, its differentiation from the surrounding tissue, its echomorphology, and specific ultrasound phenomena such as dorsal sound amplification, sound shadow and lateral sound extinction are to be considered. The diagnosis made by the sonographer is based on these phenomena. Since differential diagnostic considerations in the pre-operative assessment of pigmented and non-pigmented tumors are necessary, malignant melanoma, basal cell carcinomas, and squamous cell carcinomas were examined sonographically and evaluated by computerized analysis. Seventy-one histologically proven skin tumors were studied utilizing a computer program developed in our laboratory. In addition, we provide data as to the potential of the method.