Development and evaluation of a robust algorithm for computer-assisted detection of sentinel lymph node micrometastases Gina M Clarke, 1 Chris Peressotti, 1 Claire M B Holloway, 2,3 Judit T Zubovits, 4,5 Kela Liu 1 & Martin J Yaffe 1,6,7 1 Imaging Research, Sunnybrook Health Sciences Centre, 2 Department of Surgical Oncology, Sunnybrook Health Sciences Centre, 3 Department of Surgery, University of Toronto, 4 Department of Anatomic Pathology, Sunnybrook Health Sciences Centre, 5 Department of Laboratory Medicine and Pathobiology, University of Toronto, 6 Departments of Medical Biophysics, and 7 Medical Imaging, University of Toronto, Toronto, ON, Canada Date of submission 3 September 2010 Accepted for publication 23 November 2010 Clarke G M, Peressotti C, Holloway C M B, Zubovits J T, Liu K & Yaffe M J (2011) Histopathology 59, 116–128 Development and evaluation of a robust algorithm for computer-assisted detection of sentinel lymph node micrometastases Aims: Increasing the sectioning rate for breast sentinel lymph nodes can increase the likelihood of detecting micrometastases. To make serial sectioning feasible, we have developed an algorithm for computer-assisted detection (CAD) with digitized lymph node sections. Methods and results: K-means clustering assigned im- age pixels to one of four areas in a colourspace (representing tumour, unstained background, counter- stained background and microtomy artefacts). Four filters then removed ‘false-positive’ pixels from the tumour cluster. A set of 43 sections containing tumour (a total of 259 foci) and 59 sections negative for malignancy was defined by two pathologists, using light microscopy, and CAD was applied. For the clinically relevant task of identifying the largest focus in each section (micrometastasis in 22 ⁄ 43 sections), the sensitivity and specificity were 100%. Isolated tumour cells (ITCs) were identified in one slide initially considered to be negative. Identification of all 259 foci yielded sensitivities of 57.5% for ITCs (<0.200 mm), 89.5% for micrometastases, and 100% for larger metastases, with one false-positive. Reduced sensitivity was ascribed to variable staining. Nine additional metastases (<0.01–0.3 mm) that were not initially identified were detected by CAD. Conclusions: This algorithm is well suited to the task of sentinel lymph node evaluation and may enhance the detection of occult micrometastases. Keywords: anti-cytokeratin immunohistochemistry, breast cancer, computer-assisted detection, sentinel lymph node micrometastases, serial sectioning Abbreviations: AEC, aminoethyl carbazole; ALND, axillary node dissection; CAD, computer-assisted detection; CD, colour distance; FP, false-positive; HSV, hue-saturation value; IHC, immunohistochemistry; ITCs, isolated tumour cells; LM, light microscopy; RGB, red–green–blue; SLN, sentinel lymph node; SLNB, sentinel lymph node biopsy Introduction Disease status in the axillary lymph nodes is the most important prognostic factor for breast cancer patients. 1,2 Nodal status is incorporated into tumour stage in the Tumour Node Metastasis (TNM) classifica- tion. 3,4 Nodal status assessment has been traditionally based on axillary lymph node dissection (ALND) and, more recently, sentinel lymph node (SLN) biopsy (SLNB). 5–8 SLNB is a less invasive alternative to ALND, 9 based on the concept that there is an obligatory pathway for metastatic deposits through a limited number (1–4) Address for correspondence: G M Clarke, Sunnybrook Health Sciences Centre, 2075 Bayview Avenue, Room S6-32, Toronto, ON M4N 3M5, Canada. e-mail: gina.clarke@sunnybrook.ca Ó 2011 Blackwell Publishing Limited. Histopathology 2011, 59, 116–128. DOI: 10.1111/j.1365-2559.2011.03896.x