Phase classification of mitotic events using selective dictionary learning for stem cell populations

Abstract

Nowadays, thanks to the use of advanced technological tools, stem cell studies which play an important role in regenerative medicine and cancer studies have increased considerably. In this study, selective dictionary learning method is presented for detecting mitotic event phases in stem cells using phase contrast time-lapse microscopy images. In the proposed method, three phases are defined for representation of mitotic events. Creating a dictionary that represents these phases with a single feature space restricts the success. For this reason, three dictionaries with different features are created. Although the multiplication of image alpha values with all generated dictionaries is quite suitable for determining the lowest error value, this process is time consuming. For this reason, a selective dictionary approach based on the automatic selection of the best values with a cooperation between the dictionaries has been proposed. In this way, the high success rate is maintained and the processing time is significantly reduced. The proposed method gives better results than other state-of-art studies in terms of computational efficiency and accuracy in experiments with C2C12 and BAEC datasets.