A robust and automated cell counting method in quantification of digital breast cancer immunohistochemistry images

Quantitative analysis of immunohistochemically stained breast cancer specimens by cell counting is important for prognosis and treatment planning. This paper presents a robust, accurate, and novel method to label immunopositive and immunonegative cells automatically. During preprocessing, we developed an adaptive method to correct the colour aberration caused by imaging conditions. Next, a pixel-level segmentation was performed on preprocessed images using a support vector machine with a radial basis function kernel in HSV colour space. The segmentation result was processed by mathematical morphology operations to correct error-segmented regions and extract the marker for each cell. Validation studies showed that the automated cell-counting method had divergences varying from -5.05% to 3.99% compared with manual counting by a pathologist, indicating considerable agreement of the present automated cell counting method with manual counting. Thus, this method can free pathologists from laborious work and can potentially improve the accuracy and the reproducibility of diagnosis.