[Improvement of transrectal ultrasound. Artificial neural network analysis (ANNA) in detection and staging of prostatic carcinoma].

As a result of the enhanced clinical application of prostate specific antigen (PSA), an increasing number of men are becoming candidates for prostate cancer work-up. A high PSA value over 20 ng/ml is a good indicator of the presence of prostate cancer, but within the range of 4-10 ng/ml, it is rather unreliable. Even more alarming is the fact that prostate cancer has been found in 12-37% of patients with a "normal" PSA value of under 4 ng/ml (Hybritech). While PSA is capable of indicating a statistical risk of prostate cancer in a defined patient population, it is not able to localize cancer within the prostate gland or guide a biopsy needle to a suspicious area. This necessitates an additional effective diagnostic technique that is able to localize or rule out a malignant growth within the prostate. The methods available for the detection of these prostate cancers are digital rectal examination (DRE) and Transrectal ultrasound (TRUS). DRE is not suitable for early detection, as about 70% of the palpable malignancies have already spread beyond the prostate. The classic problem of visual interpretation of TRUS images is that hypoechoic areas suspicious for cancer may be either normal or cancerous histologically. Moreover, about 25% of all cancers have been found to be isoechoic and therefore not distinguishable from normal-appearing areas. None of the current biopsy or imaging techniques are able to cope with this dilemma. Artificial neural networks (ANN) are complex nonlinear computational models, designed much like the neuronal organization of a brain. These networks are able to model complicated biologic relationships without making assumptions based on conventional statistical distributions. Applications in Medicine and Urology have been promising. One example of such an application will be discussed in detail: A new method of Artificial Neural Network Analysis (ANNA) was employed in an attempt to obtain existing subvisual information, other than the gray scale, from conventional TRUS and to improve the accuracy of prostate cancer identification.