Formulating an approach to develop a system for the temporal analysis of clinical trial data: The POSCH AI project

The architecture and requirements of a prototypical system to help in the statistical investigation of causal relationships in a large clinical trial database are discussed. The suitability of techniques from artificial intelligence (AI), statistical analysis, and neural networks are being studied. Back-propagation experiments indicate that it works about as well as other methods but it is doubtful how useful it will be for discovering relationships in the POSCH dataset. Structure discovery techniques from artificial intelligence are being investigated.