An efficient method for automated segmentation of histochemically stained slides

Pathologists are often required to classify large numbers of histological slides. Manual classification is tedious and often not quantitatively accurate. In this paper, an automated system for the classification of histochemically stained tissue slide images using three-dimensional histograms is presented. Color space invariance of tissue clusters was used in the automation. HSV color space was used to reduce correlation between the color features. Previous approaches used shape and texture to augment color segmentation; however, these methods were computationally expensive and used empirical rules for classification. Thus, they were less than ideal. The classification technique presented overcomes these problems. As the goal of the project was batch processing of large numbers of slides, the user interaction was minimized. Images obtained from automatically classifying a set of coronary artery slides were compared with the same set done by manual classification. Comparison showed automated classification to be more accurate and reproducible