Indirect immunofluorescence is the recommended method to detect autoantibodies in patient serum, which could cause connective tissue diseases. One step of the diagnostic procedure needs to distinguish between mitotic and interphase cells in HEp-2. Furthermore, information on mitotic cell staining, namely positive and negative, could be used to improve the ability to discriminate between staining patterns of interphase cells. Despite the growing efforts to develop computer-aided-diagnosis systems supporting the diagnostic procedure, mitotic cell recognition has received little attention. Indeed this task is made difficult since the traditional classification algorithms cannot cope with the high imbalance between the number of samples in the mitotic and interphase classes. In this paper we present a classification system employing a cascade structure that first discriminates between mitotic and interphase cells tackling class imbalance and then it recognizes the positive or negative staining of a mitotic cell. It uses a set of features specifically tailored for the task at hand. The approach has been successfully evaluated on a public reference dataset of HEp-2 cells applying eleven different classification paradigms and four classifiers. The results suggest that the proposed classification system can be used to recognize mitotic cells.
Mitotic cells recognition in HEp-2 images
Iannello G;Soda P;
2014-01-01
Abstract
Indirect immunofluorescence is the recommended method to detect autoantibodies in patient serum, which could cause connective tissue diseases. One step of the diagnostic procedure needs to distinguish between mitotic and interphase cells in HEp-2. Furthermore, information on mitotic cell staining, namely positive and negative, could be used to improve the ability to discriminate between staining patterns of interphase cells. Despite the growing efforts to develop computer-aided-diagnosis systems supporting the diagnostic procedure, mitotic cell recognition has received little attention. Indeed this task is made difficult since the traditional classification algorithms cannot cope with the high imbalance between the number of samples in the mitotic and interphase classes. In this paper we present a classification system employing a cascade structure that first discriminates between mitotic and interphase cells tackling class imbalance and then it recognizes the positive or negative staining of a mitotic cell. It uses a set of features specifically tailored for the task at hand. The approach has been successfully evaluated on a public reference dataset of HEp-2 cells applying eleven different classification paradigms and four classifiers. The results suggest that the proposed classification system can be used to recognize mitotic cells.I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.