The use of X‐ray interaction data to differentiate malignant from normal breast tissue at surgical margins and biopsy analysis

X‐ray interaction data, including measuring bio‐metal levels and scattering characteristics, are being shown to be a possible discriminating variable in the classification of human tissues. However, a major concern when using X‐ray interaction data in breast cancer material is that the samples are rarely 100% tumour because of the invasive nature of the disease. The work reported here includes a methodology to help overcome this limitation as the experimental protocol includes mapping the data to histological analysis of the measured samples. This work has shown how important it is to relate the measured X‐ray parameters to the histology of the samples, particularly the clinical information that describes the percentage of tumour within each sample. Levels of K, Ca, Zn, Fe, Cu, Br and Rb were evaluated using X‐ray fluorescence and compared between tumour breast tissue and normal surrounding breast tissue. The coherent scattering properties of each sample were also examined using an angular dispersive X‐ray diffraction technique. Multivariate modelling using soft independent modelling of class analogy was used to classify samples kept out of the modelling procedure. A significant increase (p < 0.01) in the levels of Rb, Zn and K was found in the tumour samples. The levels of these elements show a correlation with the percentage of tumour reported to be present in a given sample. The results of classifying unknown tissue samples are presented using two‐class and three‐class models that help to reveal the importance of sample histology in studies involving breast cancers. Copyright © 2013 John Wiley & Sons, Ltd.