A space-occupying lesion automatic quantiication from abdominal contrast- enhanced computerized tomography images

Space-occupying lesions represent a health high risk of subjects affected by this kind of pathology. From a medical point of view, the volume occupied by each of these lesions con- stitutes the most important descriptor when addressing them, and especially for the respective decision-making process that guides their control, mitigation or elimination. In such context, this paper proposes a strategy based on computeraided image processing techniques to extract the three- dimensional morphology of a spaceoccupying lesion, of the amoebic liver abscess type, and calculate its volume. In this sense, in order to attenuate poissonian noise and improve the abscess edge information, the abdominal contrast com- puted tomography images are preprocessed using a Gauss- ian flter, an edge detector and a median flter, sequentially. Then, a clustering algorithm based on region growing procedure is applied to the enhanced images, obtaining the space- occupying lesion three-dimensional shape. Additionally, the Dice coeffcient is considered as a metric to establish the correlation between the shapes, automatic and manual lesion, the latter described by a mastologist. Then, in order to characterize the liver abscess, its volume is quantifed con- sidering both the voxels occupied by the lesion obtained by applying of the computeraided image processing, and the physical dimensions of the voxel. Finally, the automatically calculated volume is compared to that generated manually by the medical specialist. The results reveal an excellent cor- respondence between manual results and those produced by the proposed technique. This type of technique can be used as a resource not only to obtain, precisely, the value of the aforementioned descriptor, but also to monitor the process of the abscess evolution by means imaging control.