Background and Aims: Breast cancer is a leading cause of female mortality in the Western world. It is well established that the spread of breast cancer, first locally and later distally, is the major factor in patient prognosis. Experimental systems of breast cancer rely on cell lines usually derived from primary tumours or pleural effusions. Two major obstacles hinder this research: (i) Some known sub-types of breast cancers are not represented within current cell line collections; (ii) the influence of the tumour microenvironment is not usually taken into account. Experimental Design: We developed a three-dimensional assay prepared from freshly harvested breast cancer tissue embedded in soft rat collagen I cushions. Invasive behaviour and tumour response to tamoxifen therapy was measured. Changes in proliferation, apoptosis and tumour volume in response to tamoxifen treatment were quantified using image analysis software and optical projection tomography. Further cell line based experiments and histopathological analysis of resection specimens were subsequently investigated to investigate the role of EGFR signalling pathways in driving invasion. Results: We demonstrate a technique to culture primary breast cancer specimens of all sub-types. Within 2-3 weeks, individual and collective invasion of epithelial cells into the surrounding collagen I was observed using phase contrast light microscopy and histopathological methods. Addition of tamoxifen to preparations derived from ER+ tumours demonstrated a range of response as measured by proliferative and apoptotic markers and significant reduction in tumour volume not seen in ER-specimens. Changes in tumour volume allowed stratification into responsive and nonresponsive tumours. EGF within the culture medium appeared to drive a change in phenotype from ER+ to triple negative phenotype and acted as a driver for epithelial invasiveness. Conclusion: Here, we developed an assay to culture human breast tumours without sub-type bias and to investigate and quantify the spread of these ex vivo. This method could be used to quantify drug sensitivity in primary cancers under conditions closer to real life. This may provide a more predictive model than currently used cell lines.