Design of a Classification Strategy for Light Microscopy Images of the Human Liver

Light Microscopy (LM) represents the method by which pathologists study histological sections; the observations by LM can be considered the gold standard for making diagnosis and for its diagnostic accuracy. The classes that can be defined through the observation of LM images of the liver are: normal, steatosis, fibrosis, cirrhosis and hepatocarcinoma (HCC). Normally, a pathologist has to examine by LM many histological sections to perform a complete and accurate diagnosis. For this reason, an automatic system for the analysis of LM images of the liver would be particularly useful. Goal of this paper is to propose an automatic multi-stage procedure to classify the normal tissue, and the pathologic ones from human liver microphotographs. Due to the articulated nature of the examined images, the analysis will first assess if steatosis is present, by using objects analysis, and then determine whether the image belongs to a normal tissue or to one of the other pathologic ones, by using a machine learning based technique. To this aim some texture features are calculated, and the Principal Component Analysis is applied to derive the best representation of the data. Four binary Support Vector Machines classifiers are trained, one for each kind the four classes of liver conditions to be identified. Experimental results show the classification capability of the proposed system, with promising theoretical and experimental basis for developing a fully automatic decision support system.