提出了一种基于细胞神经网络（CNN）的骨髓切片有核细胞边缘检测的新方法。该方法运用粒子群优化算法设计CNN模板，利用CNN对骨髓有核细胞进行边缘检测。为了避免粒子群算法的早熟，通过判断群体适应度，在早熟种群中引入混沌寻优，增强全局寻优能力。针对有核细胞的特点，采用三级由浅入深分步学习策略，从而得到最优的CNN模板。实验结果证明了该方法的有效性，有核细胞的检出率和边缘符合度都优于传统算法。

In this paper, a novel edge detection algorithm based on cellular neural networks (CNN) is proposed for nucleated cell detection. This new algorithm applies particle swarm optimization (PSO) to design the CNN templates to identify the edge of a nucleated cell. In order to overcome the premature phenomenon of PSO, the variance of populations fitness is calculated, and chaos optimization theory is applied to enhance the PSO’s global optimization. According to the characteristics of nucleated cell, a three-step study strategy is specially designed to obtain the best CNN templates. Experimental results show the new algorithm is effective; its edge fitness rates and checkout rates are better than former algorithms.