医学X光图像感兴趣区域提取

发布时间：2018-05-05 05:45

本文选题：医学X光图像 + 感兴趣区域　；参考：《西安工业大学》2017年硕士论文

【摘要】：随着计算机辅助医疗技术的发展,医学X光图像分析逐渐成为生物医学领域的研究热点。椎体部分作为腰椎X光图像的感兴趣区域,从腰椎图像中提取椎体的信息对于腰椎疾病的诊断具有参考价值。但是受原始X光图像中存在的对比度低、边缘模糊与噪音污染等干扰因素影响,腰椎图像中椎体信息的分析面临诸多挑战。论文针对腰椎X光图像中感兴趣区域的提取进行研究,从X光图像特性与医学图像处理出发,对图像处理算法在医学X光图像处理方面的不足进行了深入的分析,提出了以去雾霾算法思想改进的反锐化掩膜增强方法、基于相位一致性的综合边缘检测方法和基于自定义广度霍夫变换的医学X光图像感兴趣区域提取方法,实现了对腰椎图像中感兴趣区域的有效提取。论文的主要工作如下:第一,对X光像的成像原理进行了深入的分析。在对X光图像充分了解的基础上对医学X光图像进行增强处理,对空间域与频率域的图像增强方法进行了学习实验,对空间域中的直方图均衡化,限定对比度自适应直方图均衡化、平滑滤波、锐化滤波、灰度变换等算法,以及频率域中的傅里叶变换、高通、低通滤波算法进行了研究,在此基础上根据雾霾图像与医学X光图像拥有的特性,引入了去雾理论的引导图滤波方法,设计实现了图像预处理的框架,良好地保持了图像边缘特性。第二,对图像的边缘提取算法进行了研究。通过实验认真研究了Sobel算子、Roberts算子、Prewitt算子、Log算子和Canny算子,在此基础上实现了基于相位一致性的综合边缘检测方法,作为自定义广义霍夫变换的边缘提取方案。第三,对感兴趣区域进行了判定与划分。针对感兴趣区域提取进行研究,实现了自定义广义霍夫变换对X光医学图像感兴趣区域提取,广义霍夫变换成功的关键在于使用相匹配的边缘信息进行分割。对大量的X光腰椎图像进行实验,实现有效的分割。论文将雾霾图像与医学X光图像处理技术相结合,在对医学X光图像进行增强的同时,保持了图像的边缘细节信息。对提取到的边缘进行区域生长,能够显示出整个椎体并且最大程度地减少噪声。自定义腰椎模板可以有效实现对腰椎X光图像的感兴趣区域进行粗粒度与细粒度的提取,为医生对于病情的诊断提供更加可靠的依据。
[Abstract]:With the development of computer aided medical technology, medical X optical image analysis has gradually become a research hotspot in the field of biomedicine. The vertebral part is a region of interest in the X light image of the lumbar spine. The information extracted from the lumbar image is of reference value for the diagnosis of lumbar disease. But the contrast of the original X light image is low. The analysis of the vertebral information in the lumbar image faces many challenges. The thesis focuses on the extraction of the region of interest in the X light image of the lumbar spine. From the characteristics of the X optical image and the medical image processing, the deficiency of the image processing algorithm in the medical X optical image processing is deeply studied. In this paper, an improved method of anti sharpening mask enhancement based on the idea of haze removal is proposed. A comprehensive edge detection method based on phase consistency and an area of interest region extraction based on medical X optical image based on the custom breadth Hof transform are used to achieve the effective extraction of the region of interest in the lumbar image. The main work of this paper is as follows: First, the imaging principle of X optical image is deeply analyzed. Based on the full understanding of the X optical image, the medical X optical image is enhanced and the image enhancement method in the space domain and frequency domain is studied. The histogram equalization in the spatial domain is balanced, the adaptive histogram equalization, smoothing filtering and sharpening are limited. Filtering, gray transformation and other algorithms, as well as Fourier transform, high pass and low pass filtering algorithms in frequency domain are studied. Based on the characteristics of haze images and medical X optical images, a bootstrap filter method of fog theory is introduced. The frame of image preprocessing is designed and realized, and the edge characteristics of the image are kept well. Second, the edge extraction algorithm of image is studied. Through the experiment, Sobel operator, Roberts operator, Prewitt operator, Log operator and Canny operator are carefully studied. On this basis, a comprehensive edge detection method based on phase consistency is implemented as the edge extraction scheme of the custom generalized Hof transform. Third, region of interest to the region of interest. According to the study of the region of interest extraction, a custom generalized Hof transform (user-defined generalized Hof transform) is used to extract the region of interest in the medical image. The key to the success of the generalized Hof transform is to use the matched edge information to divide the image. Experiments on a large number of X optical lumbar images are carried out to achieve effective segmentation. The haze image is combined with the medical X optical image processing technology to enhance the image edge details while enhancing the medical X light image. The region grows on the extracted edge, can display the whole vertebral body and reduce the noise to the maximum extent. The custom lumbar template can effectively realize the feeling of the X light image of the lumbar spine. The interesting area is coarse grained and fine-grained extraction, which provides a more reliable basis for doctors to diagnose the disease.

【学位授予单位】：西安工业大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：R814;TP391.41

【参考文献】