Tissue-characterization of the prostate using radio frequencyultrasonic signals

摘要

In this paper, we will present a complete method and system fornthe detection of prostatic carcinoma, providing color-coded images ofnthe estimated probability of malignancy by processing radio-frequencynultrasonic echo signals. For this, a hardware setup based on anconventional diagnostic sonograph was realized. The image-processingnsoftware works on ultrasound images automatically segmented into regionsnof about 3×3.5 mm. System-dependent effects, as well as tissuenattenuation, were measured and compensated for. Tissue-characterisationnparameters, which have been used successfully by other authors, werencalculated for each segment. To demonstrate the methods of selection ofnrelevant parameters and comparison of different classifiers, a firstnclinical study using data of 33 patients with local prostatic carcinomanwas performed. For these patients, location and extent of the carcinomanwere known from histological findings after radical prostatectomy.nClassifiers investigated during the study were: the linear and quadraticnBayes classifier, a nearest neighbor classifier, and several classifiersnbased on Kohonen-maps. The best classifier was used to calculatencolor-coded result images. Applying a threshold of 50% to the estimatednprobability of malignancy, produced the encouraging results of 82 andn88% for sensitivity and specificity, respectively