A Multi-Cell State Estimator Based on Contour Pheromone Field Prediction and Update

Abstract

The estimate of cell morphological parameter provides a fundamental basis for the diagnosis of diseases. Contour, as a key element of morphology, encapsulates a large number of details during cell growth, which assists biologist in judging the pathological state of cells. With this application in mind, we propose a novel multi-cell tracking algorithm fully based on the prediction and update of contour pheromone field. To accelerate the formation of pheromone field at the current frame, the pheromone field in the previous frame is predicted and obtained by a general linear model in the current frame. During the update, to ensure that the pheromone field keeps in consistence with the true cell contour, a gray-scale variance based decision model is developed and a gray gradient based model of pheromone propagation are designed as well. Experiment results show that the proposed approach can accurately and automatically estimate cell contours, and shows the superiority compared with other methods.