多模态医学图像融合技术研究

发布时间：2019-04-21 19:36

【摘要】：现如今,人们生活水平不断提高,自我保健意识逐渐增强,对自身健康也有了更多的需求。随着医疗事业如日中天的发展,医疗影像设备不断的更新换代,出现了适用于不同设备的二维医学图像。如,CT图像能清晰的显示人身体内各种脏器的解剖结构,PET图像能清晰显示人身体内功能结构。充分利用医学图像融合技术,将“功能显像”与“解剖显像”结合起来,实现优势互补。传统的医学图像融合技术虽然取得了很大的成功,但还是存在许多问题:融合规则没有统一性,针对不同融合对象的算法有差异等等。本文主要针对医学图像中脑部的CT和MRI、MRI和SPECT、MRI和PET三组图像经过多种算法和本文算法进行融合比较。为了建立图像融合的框架,本文分析了医学图像的成像原理和特点,概括性的总结了图像融合的步骤。同时,针对融合图像层次不同的特点将评价标准分成两大类。最后,本文重点分析了基于小波变换的医学图像融合算法和基于Curvelet变换的医学图像融合算法,针对基于小波变换融合后的图像无冗余信息量,但分解过程具有方向局限性,以及基于Curvelet变换的图像融合算法具有多方向性的特点,在小波变换的基础上将低频子带采用加权平均的算法,高频子带采用基于Curvelet变换的区域能量取最大的算法。实验选取10组连续的脑部MRI和SPECT图像以及10组连续的MRI和PET图像进行融合,实验结果通过客观评价标准得出一组平均值。再将本文算法的融合结果与其他算法通过7种客观评价标准进行分析比较。通过分析结果,本文的算法达到了预期效果,经本文算法融合后的图像清晰,集中了源图像的大部分能量,纹理细节反差小,无失真现象,综合评价优于其他算法。在临床医学诊断中,对病灶的定位、观测以及治疗方案的制定有重要的意义。
[Abstract]:Now, people's living standards are improving, self-care awareness is gradually enhanced, and there is more demand for their own health. With the development of medical industry, medical imaging equipment has been constantly updated, and two-dimensional medical images suitable for different equipment appear. For example, CT images can clearly display the anatomical structures of various organs in the human body, and PET images can clearly display the functional structures of the human body. Make full use of medical image fusion technology, "functional imaging" and "anatomical imaging" combined to achieve complementary advantages. Although the traditional medical image fusion technology has achieved great success, there are still many problems: the fusion rules are not uniform, the algorithms for different fusion objects are different, and so on. In this paper, CT, MRI,MRI and SPECT,MRI and PET images of the brain in medical images are fused and compared with each other. In order to establish the frame of image fusion, the imaging principle and characteristics of medical image are analyzed, and the steps of image fusion are summarized. At the same time, the evaluation criteria are divided into two categories according to the different characteristics of fusion image levels. Finally, this paper focuses on the analysis of medical image fusion algorithm based on wavelet transform and medical image fusion algorithm based on Curvelet transform. There is no redundant information in image fusion based on wavelet transform, but the decomposition process has directional limitation. The image fusion algorithm based on Curvelet transform has the characteristics of multi-directivity. On the basis of wavelet transform, the weighted average algorithm is used for low frequency subbands and the maximum region energy algorithm based on Curvelet transform is used for high frequency sub-bands. Ten consecutive brain MRI and SPECT images and 10 consecutive MRI and PET images were selected for fusion. The average values were obtained by objective evaluation criteria. Then the fusion results of this algorithm are analyzed and compared with other algorithms through 7 objective evaluation criteria. Through the analysis result, the algorithm of this paper achieves the expected effect. The fusion image of this paper is clear, concentrating most of the energy of the source image, small contrast of texture detail, no distortion phenomenon, and the comprehensive evaluation is better than other algorithms. In the clinical diagnosis, it is of great significance for the localization, observation and treatment of the lesions.
【学位授予单位】：哈尔滨工程大学
【学位级别】：硕士
【学位授予年份】：2016
【分类号】：R318;TP391.41

【参考文献】