Development of a neural network derived index for early detection of prostate cancer

摘要

ProstAsure is a neural network-derived algorithm which analyzes the profile of multiple serum tumor markers and produces a single-valued diagnostic index (ProstAsure Index, or PI) for early detection of prostate cancer (CaP). PI has been validated through multiple clinical studies with a fairly large number of blind independent test patients and has become the first of such tests commercially available through reference laboratories in the US and other countries as a clinical information processing service. We first describe briefly the development of the PI algorithm with a summary of clinical study results comparing PI with the currently accepted CaP detection tools. We then focus the discussion on two issues in developing a neural network-based clinical diagnostic system: 1) constructing training datasets under clinical constraints; and 2) estimating generalization performance by gauging the shape and "smoothness" of decision boundary surfaces of a derived classification system.