Evaluation of level set-based histology image segmentation using geometric region criteria

There is a great deal of interest in developing automated histological grading of tissue biopsies. Current approaches involve sophisticated algorithms for image segmentation, tissue architecture characterization, global texture feature extraction, and high-dimensional clustering and classification algorithms. Although overall image classification accuracy is measured, there has been very little attention paid to the quantitative assessment of the image segmentation stage (glandular structure characterization stage) to provide feedback to the segmentation process. We describe a robust approach for tissue segmentation combining spatial clustering with multiphase vector level set active contours to extract nuclei, lumen and epithelial cytoplasm. Quantitative segmentation performance compared to manual ground truth is assessed using region-based geometric criteria.