Preoperative Nomograms and Artificial Neural Networks (ANNs) for Identification of Surgical Candidates

Abstract Selection of prostate cancer risk patients for surgical treatment has traditionally been accomplished by the creation of risk groups, like clinical stage, prostate specific antigen and others. Using these data knowledge-based expert systems were created. Among these the most popular model is the logistic regression model. Ideally, this prediction should be as accurate as possible. Many studies have shown that even expert on its field often are incorrect compared to the validated nomograms and artifical neural networks (ANNs) presented herein. Nomograms are instruments that predict outcomes for the individual patient using algorithms that incorporate multiple variables. Nomograms consist of a set of axes. Each variable is represented by a scale, with each value of that variable corresponding to a specific number of points according to its prognostic impact. In a final pair of axes, the total point value from all he variables is converted to the probability of reaching the end point By using scales, nomograms calculate the continuous probability of a certain outcome, resulting in more accurate predictions than models based on risk grouping. ANNs has gained increasing popularity and are the most popular artificial learning tool in biotechnology. This technique can roughly be described as a universal algebraic function that will distinguish dependency between dependent and independent variables, which is either unknown or very complex. The application of ANNs to complex relationships makes them highly attractive for the study of complexed medical decisions like predicting pathological stage or local recurrence after radical prostatectomy (RPE). Accuracy of nomograms and ANNs for pathological staging and PSA recurrence varies between 72–88.3% versus 77–91%, and 75–81% and 67–83%, respectively.