Visualisation of Multidimensional Data for Medical Decision Support

Medical decision support tools are not widely used by clinicians, perhaps because most do not explain the decisions. We describe an approach for case-based systems using automated pattern recognition techniques. Multivariate methods estimate the degree of similarity between a new case and those in the database, and graphical displays allow users to combine this information with their own expertise. The approach is demonstrated by an example, the Spectra Visualizer, which allows radiologists to interpret magnetic resonance spectra. The variables in the overview are derived directly from the signals obtained from the scanner. Spectra with similar classification profiles can be linked to clinical history, images and expert commentary