Francesco Gentile1, Evelina La Civita2, Bartolomeo Della Ventura3, Matteo Ferro4, Michele Cennamo5, Dario Bruzzese6, Felice Crocetto7, Raffaele Velotta3, Daniela Terracciano8. 1. Nanotechnology Research Center, Department of Experimental and Clinical Medicine, University Magna Graecia of Catanzaro, Catanzaro, Italy; ElicaDea, Spinoff of Federico II University, Naples, Italy. Electronic address: francesco.gentile@unicz.it. 2. Department of Translational Medical Sciences, University of Naples "Federico II", Naples, Italy; ElicaDea, Spinoff of Federico II University, Naples, Italy. 3. Department of Physics "Ettore Pancini", University of Naples "Federico II", Naples, Italy; ElicaDea, Spinoff of Federico II University, Naples, Italy. 4. Division of Urology, European Institute of Oncology (IEO), IRCCS, Milan, Italy; ElicaDea, Spinoff of Federico II University, Naples, Italy. 5. Department of Translational Medical Sciences, University of Naples "Federico II", Naples, Italy. 6. Department of Public Health, Federico II University of Naples, Naples, Italy; ElicaDea, Spinoff of Federico II University, Naples, Italy. 7. Department of Neurosciences, Sciences of Reproduction and Odontostomatology, University of Naples Federico II, Naples, Italy. 8. Department of Translational Medical Sciences, University of Naples "Federico II", Naples, Italy; ElicaDea, Spinoff of Federico II University, Naples, Italy. Electronic address: daniela.terracciano@unina.it.
Abstract
BACKGROUND: The widespread use of prostate specific antigen (PSA) caused high rate of overdiagnosis. Overdiagnosis leads to unnecessary definitive treatments of prostate cancer (PCa) with detrimental side effects, such as erectile dysfunction and incontinence. The aim of this study was to evaluate the feasibility of an artificial neural network-based approach to develop a combinatorial model including prostate health index (PHI) and multiparametric magnetic resonance (mpMRI) to recognize clinically significant PCa at initial diagnosis. METHODS: To this aim we prospectively enrolled 177 PCa patients who underwent radical prostatectomy and had received PHI tests and mpMRI before surgery. We used artificial neural network to develop models that can identify aggressive PCa efficiently. The model receives as an input PHI plus PI-RADS score. RESULTS: The output of the model is an estimate of the presence of a low or high Gleason score. After training on a dataset of 135 samples and optimization of the variables, the model achieved values of sensitivity as high as 80% and 68% specificity. CONCLUSIONS: Our preliminary study suggests that combining mpMRI and PHI may help to better estimate the risk category of PCa at initial diagnosis, allowing a personalized treatment approach. The efficiency of the method can be improved even further by training the model on larger datasets.
BACKGROUND: The widespread use of prostate specific antigen (PSA) caused high rate of overdiagnosis. Overdiagnosis leads to unnecessary definitive treatments of prostate cancer (PCa) with detrimental side effects, such as erectile dysfunction and incontinence. The aim of this study was to evaluate the feasibility of an artificial neural network-based approach to develop a combinatorial model including prostate health index (PHI) and multiparametric magnetic resonance (mpMRI) to recognize clinically significant PCa at initial diagnosis. METHODS: To this aim we prospectively enrolled 177 PCa patients who underwent radical prostatectomy and had received PHI tests and mpMRI before surgery. We used artificial neural network to develop models that can identify aggressive PCa efficiently. The model receives as an input PHI plus PI-RADS score. RESULTS: The output of the model is an estimate of the presence of a low or high Gleason score. After training on a dataset of 135 samples and optimization of the variables, the model achieved values of sensitivity as high as 80% and 68% specificity. CONCLUSIONS: Our preliminary study suggests that combining mpMRI and PHI may help to better estimate the risk category of PCa at initial diagnosis, allowing a personalized treatment approach. The efficiency of the method can be improved even further by training the model on larger datasets.