[An artificial neural network as a tool in risk evaluation of prostate cancer. Indication for biopsy with the PSA range of 2-20 microg/l].

In this prospective study covering 5.5 years we evaluated the diagnostic power of an artificial neural network (ANN) based on PSA, %fPSA, and clinical data in the PSA range 2-20 microg/l as prostate biopsy indicator. A total of 944 patients with prostate cancer or benign hyperplasia (BPH) were analyzed. The calculation of the individual patient's risk before prostate biopsy was performed at the 90% and 95% specificity and sensitivity levels within the PSA ranges 2-4, 4.1-10, and 10.1-20 microg/l. For the low PSA range 2-4 microg/l, we recommend a first time biopsy at an ANN specificity level of 95%. For PSA range 4.1-10 microg/l, we recommend a first time biopsy at an ANN sensitivity level of 95%. A rebiopsy at the PSA range 10.1-20 micro g/l should be performed based on a 95% sensitivity level. The use of an ANN at PSA 2-20 microg/l enhances the specificity and sensitivity of %fPSA by 9-39%. The application of an ANN based on %fPSA and clinical data improves the diagnostic performance compared to %fPSA only.