ABSTRACT
The natural history of patients with renal cell cancer is bizarre: many patients succumb soon after diagnosis, while others live for decades. The lack of an accurate model to predict lifespan and the occurrence of new metastases has hampered the proper selection of therapy. In this project, a neural network programming environment (neUROn) was designed so that compiled neural networks could be tailored to specific medical/urological applications. Using neUROn, neural networks were built for data sets containing lifespan and disease progression outcomes for renal cell cancer patients. After these networks were trained, the Wilks’ generalized likelihood ratio test was used to determine which input variables were significant to the network’s prediction. An inspection of the results of this statistical test yielded information relevant to the current clinical treatment of renal cell cancer.
