基于计算机神经网络的细胞诊断系统研究

On Software for Classifying and Counting Blood Cells

在 PC机上利用真彩色细胞图像和神经网络技术 ,建立了具有自学习功能的细胞诊断与分析系统 ,有效地实现了血液细胞的 3分类和白细胞的 5分类、异常细胞的检出和收集。系统还提供了特征库的建立与编辑功能 ,可不断扩大和完善系统功能。

The Fourth Military Medical University needed software in classifying and counting blood cells. Our method is presented schematically in Fig.1. We recorded colored images of blood cells and , according to the characteristics of each type of blood cell furnished us by the medical people, we used NN (neural network) technique to classify and count the blood cells in the blood sample. Our system can not only classify and count regular types of red and white blood cells but also inspect and count extraordinary blood cells. Fig.2 shows the results of colored cell images; they show that our system is effective and the rate of accurate identification reaches 94%. We emphasize that our system can improve continuously through its self-learning capability.