当前位置:首页 >> 基础医学 >>

基于神经网络的肿瘤细胞检测


摘要 肿瘤已经成为影响人类健康的一大杀手。近年来,肿瘤的治疗方法越来越多, 治疗效果也越来越好,但是肿瘤早期的诊断却是一个比较大的难题,特别是对于 肿瘤的良性与恶性的诊断仍然是一个科学难题。针对这一问题,基于 BP 神经网络 建立诊断模型,从而建立应用于乳腺癌早期诊断的计算机辅助方法,降低源于对 临床细胞形态学诊断方法经验不足时的误判率。 本文使用 BP 神经网络,肿瘤数据采用美国新墨西哥州立大学 Neuroimaging 中心提供的、由高性能光学显微镜采集的肿瘤细胞和健康组织细胞的一系列数据, 在 Matlab 环境下通过采取肿瘤患者的医学指标,建立 BP 神经网络,使用 500 组 数据对网络进行训练,并且通过调整隐层节点的数量以及参数来达到网络的快速 收敛和更高的准确率,最终使用随机的 69 组数据进行网络测试,得到较好的预测 精度。BP 网络的应用,为解决肿瘤的良性与恶性的早期诊断,给出了一定的参考 方法。 关键词:肿瘤;医学诊断;BP 神经网络 Abstract Tumor has become a major killer of human health. In recent years, there emerges more and more cancer treatment and the treatment is getting better, but the diagnosis of cancer on early stage is still a big problem, especially in the diagnosis of the benign and malignant tumors. To solve this problem, a diagnostic model is established based on BP neural networks,a computer-aided diagnostic system can be further developed for early detection of breast tumor, and finally reduce the misdiagnosis ratio resulting from the lack of experience derived from clinical cell morphological diagnostic methods. In this paper, we apply the BP neural network to deal with the medical data including tumor cells and healthy tissue cells acquired from the Neuroimaging Center of the New Mexico State University collecting through high-performance optical microscope. After establishing the BP neural network, we pick 500 samples to train the network, adjust the number of hidden layer nodes and the corresponding parameters to achieve the networks with faster converg

赞助商链接
相关文章:
更多相关标签: