Reference: Walker, M. G. & Blum, R. L. Towards Automated Discovery from Clinical Databases: the RADIX Project. 1986.
Abstract: One of the principal goals of the RADIX project is the development of programs to automate the discovery of medical knowledge from patient databases. Our first program, called RX, succeeded in replicating many known medical relationships including some findings not known to us. Our current work is proceeding in two areas: exploratory data analysis and artificial intelligence. The literature of each of these fields relevant to automated discovery is reviewed. Two ares of our current work are described: 1) computationally efficient statistical indices to detect clustering and regularity in scattergrams, and 2) knowledge-based techniques to replicate hypothesis formation from the examination of patient records. In creating disease hypotheses researchers make extensive use of notions of causation, mechanisim of action, tempo. and quantitative sufficiency, as well as detailed knowledge of pathophysiology. We prsent an example of the discovery of late- stage lupus nephropathy in which we seek to automate this process of hypothesis formation.
Notes: Journal Memo.