Indirect immunofluorescence (IIF) is the recommended method to diagnose the presence of antinuclear autoantibodies in patient serum. A main step of the diagnostic procedure requires to detect mitotic cells in the well under examination. However, such cells rarely occur in comparison to other cells and, hence, traditional recognition algorithms fail in this task since they cannot cope with large differences between the number of samples in each class, resulting in a low predictive accuracy over the minority class. In this paper we present a system for mitotic cells recognition based on multiobjective optimisation, which is able to handle their low a priori probability. It chooses between the output of a classifier trained on the original skewed distribution and the output of a classifier trained according to a learning method addressing the course of imbalanced data. This choice is driven by a parameter whose value maximises, on a validation set, two objective functions, i.e. the global accuracy and the accuracies for each class. The approach has been evaluated on an annotated dataset of mitotic cells and successfully compared to five learning methods applying four different classification paradigms.
Mitotic HEp-2 Cells Recognition under Class Skew
PERCANNELLA, Gennaro;VENTO, Mario
2011
Abstract
Indirect immunofluorescence (IIF) is the recommended method to diagnose the presence of antinuclear autoantibodies in patient serum. A main step of the diagnostic procedure requires to detect mitotic cells in the well under examination. However, such cells rarely occur in comparison to other cells and, hence, traditional recognition algorithms fail in this task since they cannot cope with large differences between the number of samples in each class, resulting in a low predictive accuracy over the minority class. In this paper we present a system for mitotic cells recognition based on multiobjective optimisation, which is able to handle their low a priori probability. It chooses between the output of a classifier trained on the original skewed distribution and the output of a classifier trained according to a learning method addressing the course of imbalanced data. This choice is driven by a parameter whose value maximises, on a validation set, two objective functions, i.e. the global accuracy and the accuracies for each class. The approach has been evaluated on an annotated dataset of mitotic cells and successfully compared to five learning methods applying four different classification paradigms.I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.