A prototype decision support system for MR spectroscopy-assisted diagnosis of brain tumours.

Our objective is to develop a decision support system that improves the accuracy of non-invasive brain tumour diagnosis and grading by enabling radiologists to use data from Magnetic Resonance Spectroscopy (MRS). The system, which uses pattern recognition techniques, is trained on a validated database of spectra and associated clinical information to provide automated classification of spectra from brain tumours. An innovative user-interface presents classification results as a two-dimensional overview plot in which points representing cases of different diseases form distinct clusters. Users can inspect any cases in these plots and compare them with the new, unknown spectrum. Hence, the overview plot can both communicate the classification of a case and help provide explanation for that classification in part by supporting human case-based reasoning. This paper describes the development of a prototype system implemented in JAVA.