基于B超脉冲回波RF信号鉴别甲状腺结节良恶性的研究

发布时间：2018-03-31 09:21

  本文选题：甲状腺结节　切入点：射频信号　出处：《南京大学》2017年硕士论文

【摘要】：随着超声技术的不断发展及人们健康意识的提高,甲状腺结节在超声检查中的发现率已达70%以上,其中恶性约占5%。当前,甲状腺恶性结节(甲状腺癌)对人类身心健康已经构成了重大威胁。医学超声成像能够表现出下列优势:无损性、可重复性好、实时性、廉价性、灵敏度高等,常常应用于甲状腺结节良恶性的鉴别中。目前,甲状腺结节良恶性鉴别的金标准是细针穿刺抽吸活检(FNAB),为减少患者的痛苦,缩短检查周期,利用B超脉冲回波RF信号中的特征量以及神经网络等信号处理技术对甲状腺结节良恶性进行判断,有助于为临床医生更准确、更全面的做出诊断提供辅助信息。本研究利用甲状腺B超脉冲回波RF信号,采用时域、频域及非线性分析方法和BP神经网络识别技术,探讨一种新的甲状腺结节良恶性鉴别方法。选取甲状腺结节感兴趣区域内的B超脉冲回波RF信号,提取多个特征量,如期望值、低频小波系数均值、小波模极大值均值和最大李雅普诺夫指数,并利用BP神经网络进行甲状腺结节识别量化。实验结果表明,以上特征量均能有效地对甲状腺结节进行良恶性鉴别,其识别率分别高达95.1%,92.7%,97.6%及82.9%。其中,期望值、低频小波系数均值、小波模极大值均值对2级甲状腺结节的识别率达 100%。本文工作为诊断甲状腺结节良恶性提供了一个新的思路,具有重要的临床应用价值。
[Abstract]:With the development of ultrasound technology and the improvement of people's health awareness, the detection rate of thyroid nodules in ultrasound examination has reached more than 70%, of which malignancy accounts for about 5%.At present, malignant thyroid nodules (thyroid carcinoma) have posed a major threat to human physical and mental health.Medical ultrasound imaging can show the following advantages: nondestructive, reproducible, real-time, cheap, high sensitivity and so on. It is often used in the differential diagnosis of benign and malignant thyroid nodules.At present, the gold standard for differentiating benign and malignant thyroid nodules is fine needle aspiration biopsy (FNABN).The diagnosis of benign and malignant thyroid nodules by using the characteristic quantity of RF signal of B-mode ultrasonic echo and neural network is helpful to provide more accurate and comprehensive diagnosis information for clinicians.In this study, a new method for differentiating benign and malignant thyroid nodules was studied by using time-domain, frequency-domain and nonlinear analysis methods and BP neural network.The RF signal of B-mode ultrasonic echo in the region of interest of thyroid nodule is selected, and several characteristic quantities are extracted, such as expectation value, mean of low frequency wavelet coefficient, mean value of wavelet modulus maximum and maximum Lyapunov exponent.BP neural network was used to identify thyroid nodules.The experimental results show that the above characteristic quantities can effectively distinguish benign and malignant thyroid nodules, and the recognition rates are as high as 95.1% and 82.9%, respectively.Among them, the expected value, the mean of low frequency wavelet coefficient and the mean of wavelet modulus maximum value are 100% for the recognition of thyroid nodule of grade 2.This work provides a new idea for the diagnosis of benign and malignant thyroid nodules and has important clinical application value.
【学位授予单位】：南京大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：R581;R445.1
，

本文编号：1690113