Benecchi L.1, Bocchi F.1, Potenzoni M.2, Russo F.1, Perucchini L.1, Quarta M.1, Tonghini M.1, Bocchi P.1, Del Boca C.1
1Istituti Ospitalieri Di Cremona, Dept. of Urology, Cremona, Italy, 2Fidenza Hospital, Dept. of Urology, Parma, Italy
INTRODUCTION & OBJECTIVES: Fuzzy system and neural network are complementary technologies in the design of adaptive intelligent system. Artificial neural network (ANN) learns from scratch by adjusting the interconnections between layers. A Neuro-fuzzy system is simply a fuzzy inference system trained by a neural network-learning algorithm. The aim of our work is to develop a Neuro-Fuzzy system to predict a positive prostate biopsy.
MATERIAL & METHODS: 
We retrospectively reviewed 1280 patients who underwent prostate biopsy. All men had a PSA level of 20 ng/ml or less. 469 men (36.6%) had prostate cancer. A neuro-fuzzy system was developed using a coactive neuro-fuzzy inference system model. The model was composed of an input layer with four neurons (PSA, percent free PSA, PSA density and age), and an output neuron representing the output value of the predictor. The cases were random divided in train-test group (800cases) and validation group (480cases).
RESULTS: In the validation group the area under the curve (AUC) for the neuro-fuzzy system output was 0.777 +/-0.024 (95% confidence interval 0.737 to 0.814), for PSA was 0.545 +/-0.028 (95% confidence interval 0.499 to 0.590) and for percent free PSA was 0.728 +/-0.023 (95% confidence interval 0.686 to 0.767). Furthermore, pairwise comparison of AUCs evidenced differences among PSA, percent free PSA and PSA density versus neuro-fuzzy system (PSA versus neuro-fuzzy system’s output, P=0.001; PSA density versus Neuro-fuzzy system'output, P=0.007; percent free PSA versus neuro-fuzzy system’s output, p=0.001).
CONCLUSIONS: We constructed a neuro-fuzzy system based on both serum data and clinical data (age, total PSA, %free PSA, and PSA density) to identify men at risk of harboring prostate cancer. It may assist patients and clinicians in deciding whether further prostatic evaluations are necessary.