基于神经网络的肿瘤细胞检测.docVIP

摘要 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 convergence and higher accuracy rate. At last, we use random 69 samples for network testing, and get good prediction accuracy. The application of the BP neural networks to the early diagnosis of the benign and malignant tumors is a good reference for future study. Key words: Cancer; Medical diagnosis; BP neural network 目录 1 绪论 1 1.1 课题背景与意义 1 1.2 课题研究现状 1 1.3 课题研究的主要内容 2 1.4 Matlab工具介绍 3 1.4 研究步骤及方法 5 2 神经网络方法 6 2.1 人工神经网络概述 6 2.1.1 人工神经网络的概念 6 2.1.2 人工神经网络的主要方向 8 2.1.3 人工神经网络特点 8 2.1.4 人工神经网络优缺点 9 2.2 径向基网络 10 2.3 支持向量机 11 2.4 小波神经网络 12 2.5 BP神经网络 12 2.5.1 BP神经网络优点 13 2.5.2 BP神经网络缺点 13 2.5.3 BP神经网络算法与步骤 14 3 研究方法 16 3.1 数据解释 16 3.2 模型建立 16 3.3 训练过程 17 4 结果 21 4.1 网络优化 21 4.2 数据实测 21 5 讨论 23 总结 24 致谢 25 参考文献 26 1 绪论 1.1 课题背景与意义 肿瘤按照最简单的分类为两种：良性肿瘤与恶性肿瘤。 顾名思义，良性肿瘤对于人类并无致命性，可通过现代医疗手段治愈，而恶性肿瘤又称癌症，乃是当今医学无法攻克的难题之一，其极高的病死率常年来都威胁着人类的生命健康。 一连串惊人的数字、例证，无情的发射出惊醒人类的信号。当今世界，对人类生命威胁最严重的疾病之一仍属癌症。在我们生活的这个地球上，每年约有500