Sequential sparse representation for mitotic event recognition

Proposed is a sequential sparsity representation method for mitotic event recognition. First, an imaging model-based microscopy image segmentation method is implemented for mitotic candidate extraction. Then, a sequential sparsity representation scheme is proposed for dictionary learning and sparsity decomposition for sequential events. Specifically, a convex objective function jointly regularised by sparsity, consistent and smooth terms is formulated to compute the reconstructed residual, which is finally utilised for classification. This method can take advantage of temporal context for spatio-temporal event modelling. Moverover, it can overcome the shortage of temporal inference models which highly depends on a large amount of training data and long-range temporal context. The comparison shows that this method can outperform competing methods in terms of precision, recall and F1 score.