SVM分类器在肝癌早期诊断中的应用研究

发布时间：2018-03-26 05:32

本文选题：肝癌　切入点：AFP　出处：《大连理工大学》2016年硕士论文

【摘要】：肝细胞癌(HCC)是世界范围内癌症相关死亡的第二大病因,在世界范围内肝癌的发病率都呈上升趋势,全球每年大约有75万的新发病例。在中国,基于人口的研究表明,肝细胞癌的发病率和死亡率在所有的癌症类型里均排在第二位,并且其发病率近似于死亡率,这说明患有肝癌的大部分患者都死于肝细胞癌。在临床上,我们通常通过检测AFP和腹部超声波检测法来诊断肝癌。而这种传统的诊断方法存在两个问题：1.当检测到AFP异常时,患者大多数往往已经到了肝癌晚期,这时候不管是手术还是放疗或化疗等,患者的治愈率都是极低的,并且花费极高；2.AFP并不是诊断肝癌的唯一标志物,有时这个标志物在诊断肝癌时是无效的,因此可能导致误诊,延误了患者的最佳治疗期。因此本篇论文以某医院提供的当地早期肝癌、肝病患者的检测指标为研究对象,提取这些指标之间的隐藏模式和关系,通过这些指标及关系构建诊断早期肝癌的分类器,从而达到尽早预测肝癌、提高诊断肝癌准确率的目的。研究内容包括以下方面：(1)分析早期肝癌患者和肝病患者的各项检测指标,并对数据进行分析和预处理。从分析结果上看,早期肝癌患者的AFP值大多数都处于正常范围内,与肝病患者相当；而这两类患者的其他检测指标大多数都高于正常范围。因此两者之间的检测指标存在交叉部分,单靠检测指标并不能将肝癌患者与肝病患者区分开来。数据预处理方面,利用关联算法提取特异性指标,利用特征选择和主成分分析(PCA)对数据进行降维处理。(2)利用SVM对肝癌患者和肝病患者进行分类研究,建立支持向量机分类器模型。此分类器模型整合了经过数据预处理得到的16种特异性指标,这些指标对于肝癌的诊断具有重要意义；同时利用网格划分法和粒子群优法算法优化SVM模型参数,最终分类器模型的预测准确率分别是94.186%和93.0233%。(3)针对基本粒子群算法容易陷入局部最优的问题,提出了自适应变异粒子群算法,利用变异操作对粒子群算法进行优化,以求达到全局最优解,提高分类器的准确率,最终得到的预测准确率是95.3488%。
[Abstract]:Hepatocellular carcinoma (HCC) is the second leading cause of cancer-related deaths worldwide, and the incidence of liver cancer is on the rise worldwide, with about 750000 new cases per year worldwide. In China, population-based studies show that. The morbidity and mortality of hepatocellular carcinoma are the second highest among all cancer types, and the incidence is similar to the mortality rate, which means that most patients with liver cancer die from hepatocellular carcinoma. We usually use AFP and abdominal ultrasound to diagnose liver cancer. But there are two problems with this traditional diagnostic method: 1. When we detect abnormal AFP, the majority of patients tend to have advanced liver cancer. At this time, whether it is surgery, radiotherapy or chemotherapy, the cure rate of patients is extremely low, and the cost of AFP is extremely high. 2. AFP is not the only marker for the diagnosis of liver cancer. Sometimes this marker is ineffective in the diagnosis of liver cancer, so it may lead to misdiagnosis. The best treatment period is delayed. Therefore, this paper takes the local early liver cancer and liver disease detection index provided by a hospital as the research object, and extracts the hidden pattern and relationship between these indexes. Based on these indexes and relationships, a classifier for early diagnosis of liver cancer was constructed to predict liver cancer as early as possible. The purpose of this study is to improve the accuracy of diagnosis of liver cancer. The contents of the study include the following aspects: 1) Analysis of the early stage liver cancer patients and liver disease patients, and analysis and preprocessing of the data. From the perspective of the analysis results, Most of the AFP values of patients with early liver cancer are within the normal range, which is similar to that of the patients with liver disease, and most of the other indexes of the two groups are higher than the normal range. The detection index alone can not distinguish the liver cancer patients from the liver disease patients. In the aspect of data preprocessing, the association algorithm is used to extract the specific indexes. Feature selection and principal component analysis (PCA) were used to reduce the dimension of the data. (2) SVM was used to classify the patients with liver cancer and liver disease. The support vector machine classifier model is established. The classifier model integrates 16 specific indexes obtained by data preprocessing, which are of great significance for the diagnosis of liver cancer. At the same time, the SVM model parameters are optimized by using mesh division method and particle swarm optimization algorithm. The prediction accuracy of the final classifier model is 94.186% and 93.02333%, respectively. An adaptive mutation particle swarm optimization algorithm is proposed, in order to achieve the global optimal solution and improve the accuracy of the classifier, the prediction accuracy is 95.3488%.
【学位授予单位】：大连理工大学
【学位级别】：硕士
【学位授予年份】：2016
【分类号】：TP18;R735.7

【参考文献】