Segmentation of nuclei in digital pathology images

There are challenges for image cancer nuclei segmentation in clinical decision support systems for brain tumor diagnosis. In this study, we propose a method for segmentation of cancer nuclei when such conflicts of cancer nuclei involve `omics' indicative of brain tumors pathologically. To constrain the problem space in the region of color information (i.e. cancer nuclei), we begin by converting the images into the V component of HSV (Hue, Saturation, Value) using the level-set segmentation (VLS) in the training stage, follow by applying the sparsity representation (SR) in the test stage. Via the SR, the proposed VLS-SR would exhibits an improved capability of searching recursively for the optimal threshold level-set in the working subsets of the SR for image cancer nuclei segmentation.