Tissue characterization in IVUS images is a crucial problem for the physicians for studying the vascular diseases. However, this task is complex and suffers from multiple drawbacks (slow manual process, subjective interpretation, etc.) Therefore, automatic plaque characterization is a highly desirable tool.
However, automatic tissue characterization is a problem of high complexity. First of all, we need a unique and powerful description of the tissues to be classified. This is done by the feature extraction process, that in order to obtain complete and meaningful description, image features should be based on texture. Thus, a study of the most representable feature spaces is done, to conclude with some enlightening results. After analyzing the experimental results, we conclude that co-occurrence matrix measures, local binary patterns, and o
Was this article helpful?