BACKGROUND: Intraoperative diagnosis of sentinel node metastases enables an immediate decision to proceed to axillary lymph node dissection, avoiding a second operation in node-positive women with breast cancer. METHODS: An optical scanner was developed that interrogated the cut surface of bivalved, but otherwise unprocessed, sentinel lymph nodes with pulses of white light by elastic scattering spectroscopy (ESS). The scattered light underwent spectral analysis, and individual spectra were initially correlated with conventional histology to develop a diagnostic algorithm. This algorithm was used to create false colour-coded maps of scans from an independent set of nodes, and the optimal criteria for discriminating between normal and cancer spectra were defined statistically. RESULTS: The discriminant algorithm was developed from a training set of 2989 spectra obtained from 30 metastatic and 331 normal nodes. Subsequent scans from 129 independent nodes were analysed. The scanner detected macrometastases (larger than 2 mm) with a sensitivity of 76 per cent (69 per cent including micrometastases) and specificity of 96 per cent. CONCLUSION: In this proof-of-principle study, the ESS results were comparable with current intraoperative diagnostic techniques of lymph node assessment.