Pooled time series representation for mitosis event recognition

Abstract

This paper proposes a new feature representation for mitotic event detection in time-lapse phase contrast microscopy image sequences of stem cell populations. First, an imaging model-based microscopy image segmentation method is implemented for mitotic candidate extraction. Then, a new feature representation framework based on time series pooling is proposed for sequential events. At last, a support vector machine classifier is utilized for mitotic cell modeling and detection. Different from other feature representations including bag-of-visual-words when using identical underlying feature descriptors, this method can take advantage of temporal relations among frames, the idea is to keep track of how descriptor values are changing over time and summarize them to represent appearance in the cell sequence. The comparison experiments demonstrate the superiority of the proposed method.