SELF-ORGANIZING MAPS FOR THE INVESTIGATION OF CLINICAL-DATA - A CASE-STUDY

The clinical process often involves comparisons of how one set of meas urements is related to previous, similar, data and the use of this inf ormation to take decisions concerning possible courses of action, ofte n with insufficient data to make meaningful calculations of probabilit ies. Self-organising maps are useful devices for data visualisation. T o illustrate how visualisation with self-organising maps might be used in the clinical process, this paper describes the investigation of an osteoporosis data set using this technique. The data set had previous ly been used to show that backpropagation neural networks were capable of distinguishing between patients who had suffered a fracture, and t hose who had not using measured bone mineral density values; illustrat ing the power of these networks to model relationships in data. Howeve r, we had realised that this was somewhat of an academic exercise give n that in reality a non-fracture case might be a fracture case waiting to happen. We felt it would be more productive to examine the data it self rather than model an imposed classification. As part of this inve stigation, the data set was examined using self-organising maps. From the results of the investigation, we conclude that it is possible to c reate a map, a compressed data representation, using BMD values which may then be partitioned into low and high fracture risk areas. Using s uch a map may be a useful screening mechanism for detecting people at risk of osteoporotic fracture.