Recognition of red cell and megakaryocyte based L-shaped envelope function

In order to explore the cell composition and its metabolism, it is important to let computer recognize the cells and get the counts of different cells for a sample. This paper proposes an L-shaped envelop function and the related fuzzy clustering method as a way to identify the megakaryocyte and the red cell from the sliced marrow image. This method is useful when the staining is insufficient and the color cannot be used as the identifying factor. This method uses the experimental histogram data to fit the L-shaped function and then use it as the envelop for the match test. The fuzzy c-means (FCM) performance index is used to test the adjacent area and get the minimum and finally secure the identification. The new method is not limited to megakaryocyte or red cell and can be used for general purposes of cell recognition. Tests show that this envelop function can ensure the recognition rate with different staining batches and can reach satisfied counting under similar illumination condition.