基于MRI图像的胰腺肿瘤鉴别诊断分析

发布时间：2018-03-14 08:42

  本文选题：meta分析　切入点：胰腺癌　出处：《电子科技大学》2014年硕士论文　论文类型：学位论文

【摘要】：胰腺癌是一种常见的恶性程度很高的胰腺肿瘤,作为全球致死率最高的疾病之一,世界各国都在加大力度研究胰腺癌的预防、诊断、预后以及治疗过程。CT在诊断胰腺癌、可切除性手术评估、胰腺癌分期方面有比较理想的效果,但是对于病灶较小的情况和较复杂的病变,CT的诊断效果就不尽如人意。MRI是CT之后在近年迅速发展起来较新的一项影像技术,在胰腺癌的诊断方面表现出更好的效果,现在很多研究者认为MRI在诊断胰腺癌时优于CT,MRI作为当前胰腺疾病诊断的主要手段之一表现出很多优点,尤其是结合DWI和MRCP技术其诊断效果超越了当前其他诊断手段。MRI和CT技术作为诊断胰腺癌最常用的两种影像技术,其诊断效果一直被人们拿来对比,各种研究得出来的结果有很多不一致的地方,甚至结论完全相反,我们针对MRI和CT在胰腺癌诊断方面整合近年的研究做一个meta分析,以做出一个新的统一的结论,为胰腺癌诊断提供指导意见。在这篇meta分析中我们主要研究几个方面的内容。首先,对近十年来的MRI和CT诊断胰腺癌的文献进行了全面的检索,严格按照纳入标准筛选所有可能相关的文献,最后纳入了一些高质量论文作为后续研究数据;其次根据数据选择和提取标准提取了研究所需的所有指标和必要数据,综合分析得出了一个统一的结论;最后我们针对MRI和CT对胰腺癌诊断可能存在的影响因素进行了亚组分析,通过统计检验计算出其影响诊断准确率的显著程度。meta分析结果显示,MRI诊断胰腺癌的敏感性和CT相仿,但是MRI的特异性要明显高于CT,诊断胰腺癌的时候MRI是个更好的诊断手段。虽然影像学高速发展给胰腺癌的诊断带来极大的方便,但是对胰腺癌准确的进行诊断依然是一项难题,为了辅助影像科医生诊断胰腺癌和减轻影像科医生的阅片工作,本文随后对胰腺癌MRI图像进行了计算机辅助诊断研究,对比了SVM和LDA方法对胰腺癌、胰腺炎患者和正常人的分类准确率。这部分工作主要包含以下流程:在医院采集患者数据,根据专家指导进行感兴趣区域提取并提取图像特征,实现机器学习算法后对训练数据集进行分类器训练,对MRI图像中胰腺疾病特征进行量化,完成人眼视觉特征到计算机数据特征的转换,最后对未知胰腺疾病患者MRI数据进行分类,为临床提供计算机辅助诊断服务。
[Abstract]:Pancreatic cancer is a common malignancy with high pancreatic cancer, as one of the highest rates of fatal diseases in the world, all the countries in the world in the diagnosis of prevention, strengthen the study of pancreatic cancer, prognosis and treatment of.CT in the diagnosis of pancreatic cancer, surgical resectability assessment, has the ideal effect of pancreatic cancer. But for the small lesions and complex lesions, the diagnostic effect of CT is not satisfactory after CT in.MRI is developed rapidly in recent years a new imaging technique, shows better effect in the diagnosis of pancreatic cancer, now many researchers believe that MRI is superior in the diagnosis of pancreatic cancer CT, MRI as one of the principal means to the diagnosis of pancreatic disease showed a lot of advantages, especially the combination of DWI and MRCP technology as the diagnostic results beyond the diagnosis of pancreatic adenocarcinoma is the most of the other diagnostic means of.MRI and CT Technology Two kinds of imaging techniques commonly used, its diagnosis result has been compared with the results of the study, all have a lot of inconsistencies, even the opposite conclusion, we focus on the MRI and CT in the diagnosis of pancreatic cancer and integration of recent research and analysis to make a meta, to make a new unified conclusion and provide guidance for the diagnosis of pancreatic cancer. In this article we mainly analyze meta several research contents. Firstly, for the past ten years, MRI and CT in the diagnosis of pancreatic cancer literature conducted a comprehensive search, in strict accordance with the inclusion criteria for screening all relevant documents, finally incorporated a number of high quality papers as the follow-up study data; secondly according to the data selection and extraction standard extract all the indicators required and necessary data, comprehensive analysis and draw a unified conclusion; finally we aimed at MRI and CT in pancreatic cancer The subgroup analysis of the factors affecting the diagnosis may exist, through statistical tests to calculate the degree of.Meta analysis and the accuracy of its diagnostic display, MRI diagnosis of pancreatic cancer and the sensitivity of CT is similar, but the specificity of MRI was higher than that of CT, when the MRI diagnosis of pancreatic cancer is a better diagnostic tool. Despite the rapid development of imaging brings great convenience for the diagnosis of pancreatic cancer, but the accurate diagnosis of pancreatic cancer is still a difficult problem, in order to assist doctors of Radiology Diagnosis of pancreatic cancer and reduce radiologists reading work, then carry out a research on computer aided diagnosis of pancreatic cancer MRI image, compared to SVM and the LDA method for pancreatic cancer, pancreatitis and normal classification accuracy. This work mainly includes the following processes: collecting patient data in the hospital, under the guidance of experts are interested in Region extraction and extraction of image features, to achieve machine learning algorithm on the training data set classifier training to quantify pancreatic disease characteristics of the MRI image, the conversion is complete the visual characteristics to the computer data characteristics, and finally classify unknown pancreatic disease MRI data, provide computer aided diagnosis for clinical service.

【学位授予单位】：电子科技大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：R445.2;R735.9
，

本文编号：1610456