Integrated data visualisation and classification using growing cell structure neural networks applied to the diagnosis of breast cancer

The technique described provides a comprehensive integrated tool for the visualisation, analysis and classification of multidimensional data that is simple and intuitive to use owing to its visual nature. As an integrated tool, it provides users with the maximum information, allowing them to see how the classification system works and which subgroups of data it is able to classify confidently and which it is not. Further, it allows the potential for users to specify arbitrary decision boundaries to discriminate between subgroups of the data of particular interest in a way that is not possible in general classification systems. The technique has been demonstrated by application to the diagnosis of breast cancer from FNAB (Fine Needle Aspirates of Breast tissue) data, where it proved highly successful both in terms of classification results and the detection of correlations between input variables and the desired classification. The promise of the technique has been further demonstrated by the reaction of experienced doctors who found the method to be both simple and intuitive to use and in some respects more useful than existing statistical techniques