Computational Vision and Medical Image Processing IV

254

lesion, micronodular cirrhosis or granuloma with

necrosis. It consisted in two steps: a segmentation

on the basis of texture properties of the image

was provided, allowing the choice of small regions

inside and outside the abnormal regions where

the successive classification step could be applied

and the labels of identified abnormal regions can

be used in order to evaluate the classification

accuracy.

The proposed procedure to classify tissues from-

microscopic liver images has been tested consider-

ing microscopic images of liver tissue, providing

encouraging results. Some difficulties are still

present when the texture is not well described by

the considered features. Therefore two further

developments are ongoing. The former concerns ...