基于超声图像分析的肝癌射频治疗定量评价
|
摘要
肝癌已成为当前人类因癌症死亡的主要原因之一,有资料表明,我国肝癌发病率将在相当时期内呈现显著上升趋势。近年来射频消融治疗方法因其具有微创、安全、对肝功能影响较小等优点,被广泛应用于肝癌的治疗方面。超声图像具有廉价、实时、无辐射及可重复性等优势,成为当前医学影像技术的一个重要分支。临床上可以采用超声图像来评价肝癌射频治疗的效果,但是超声图像具有低分辨率、固有的斑点噪声等缺点,对医生观察和分析图像造成了一定障碍。因此,国内外很多学者采用计算机辅助对超声图像进行研究并取得了一定的成果。但大部分研究都是提取图像特征对肝脏疾病确认及分类,很少有人对肝癌射频治疗后的恢复过程及治疗效果给出定量的评价。
为了实现肝癌射频治疗效果的定量评价研究,在大连医科大学动物实验中心进行了动物造模实验以获得肝脏超声图像,对不同肝脏区域的超声图像进行分类,用中值滤波的方法对图像进行预处理,以抑制超声图像特有的斑点噪声。在实验中采用了灰度共生矩阵、灰度游程长度、Gabor变换和小波分解等经典纹理特征提取方法对各类超声图像进行特征提取,并对实验结果进行研究和分析。
本文提出了基于变差函数和奇异值分解的肝癌射频治疗效果的定量评价方法。通过计算变差函数值提取图像的纹理特征,并利用该特征对图像的奇异值之和进行加权,从而得到肝癌射频治疗的定量评价体系。建立检验样本对此评价体系进行检验,正确率达到93.3%,因此可以利用这一定量评价体系辅助医生做出比较可靠的复查意见。
实验中所用到的超声图像均是由医生指定感兴趣区域,然后进行手动分割。本文进行图像分割算法的研究,期望用自动分割代替手动分割,实现癌变区域的自动检测。介绍了最大类间方差阈值分割法、最大熵阈值分割法和基于模糊划分熵等几种医学图像处理中常用的图像分割算法,并对实验结果进行了分析。
Liver cancer is one of the main causes of death. Data indicate that the number of people who has liver cancer is increasing. In recent years, radiofrequency ablation is widely used in the treatment of liver cancer because of its minimally invasive, safe, less impact on liver. Ultrasound images have the advantages of low-cost, real-time, no-radiation and repeatability, and become an important branch of medical imaging techniques. In clinical, ultrasound images can be used to evaluate the effects of radiofrequency ablation treatment for liver cancer, but its low-resolution and speckle noise are obstacle for doctors to observe and analyze the images. Therefore, many academics at home and abroad did a lot of research on ultrasound images by computer and have achieved certain results. But most studies just confirm or classify the liver disease by extracting the image feature and few people give a quantitative evaluation on the effect of radiofrequency treatment.
In order to study the effect of liver cancer radiofrequency treatment, animal model experiments are performed at Dalian Medical University to get ultrasound images.The method of median filter is applied to ultrasound image preprocessing to suppress the speckle noise. In the experiment, gray co-occurrence matrix, gray level run length matrix, Gabor transform and wavelet decomposition methods are used to extract the texture features from different ultrasound images and the experiment results and analysis are given out.
A quantitative evaluation method using variogram function and singular value decomposition is proposed in this paper. Texture features of the images are extracted from calculating the variogram function value, and then a quantitative evaluation system is obtained using the texture features to weight sum of the singular value. Test samples are used in this paper to test the reliability of the system, and the accuracy is 93.3%,so the quantitative evaluation system can be used to assist doctors to evaluating the effect of the liver cancer radiofrequency treatment more accurately.
The region of interest of the ultrasound images we get are all manual segmentation. Image segmentation algorithm is studied in this paper and we hope that we can use automatic segmentation instead of manual segmentation to achieve automatic detection of cancer. Otsu threshold method, maximum entropy threshold method and fuzzy partition entropy method are used in ultrasound image segmentation and experiment results are given out.
为了实现肝癌射频治疗效果的定量评价研究,在大连医科大学动物实验中心进行了动物造模实验以获得肝脏超声图像,对不同肝脏区域的超声图像进行分类,用中值滤波的方法对图像进行预处理,以抑制超声图像特有的斑点噪声。在实验中采用了灰度共生矩阵、灰度游程长度、Gabor变换和小波分解等经典纹理特征提取方法对各类超声图像进行特征提取,并对实验结果进行研究和分析。
本文提出了基于变差函数和奇异值分解的肝癌射频治疗效果的定量评价方法。通过计算变差函数值提取图像的纹理特征,并利用该特征对图像的奇异值之和进行加权,从而得到肝癌射频治疗的定量评价体系。建立检验样本对此评价体系进行检验,正确率达到93.3%,因此可以利用这一定量评价体系辅助医生做出比较可靠的复查意见。
实验中所用到的超声图像均是由医生指定感兴趣区域,然后进行手动分割。本文进行图像分割算法的研究,期望用自动分割代替手动分割,实现癌变区域的自动检测。介绍了最大类间方差阈值分割法、最大熵阈值分割法和基于模糊划分熵等几种医学图像处理中常用的图像分割算法,并对实验结果进行了分析。
Liver cancer is one of the main causes of death. Data indicate that the number of people who has liver cancer is increasing. In recent years, radiofrequency ablation is widely used in the treatment of liver cancer because of its minimally invasive, safe, less impact on liver. Ultrasound images have the advantages of low-cost, real-time, no-radiation and repeatability, and become an important branch of medical imaging techniques. In clinical, ultrasound images can be used to evaluate the effects of radiofrequency ablation treatment for liver cancer, but its low-resolution and speckle noise are obstacle for doctors to observe and analyze the images. Therefore, many academics at home and abroad did a lot of research on ultrasound images by computer and have achieved certain results. But most studies just confirm or classify the liver disease by extracting the image feature and few people give a quantitative evaluation on the effect of radiofrequency treatment.
In order to study the effect of liver cancer radiofrequency treatment, animal model experiments are performed at Dalian Medical University to get ultrasound images.The method of median filter is applied to ultrasound image preprocessing to suppress the speckle noise. In the experiment, gray co-occurrence matrix, gray level run length matrix, Gabor transform and wavelet decomposition methods are used to extract the texture features from different ultrasound images and the experiment results and analysis are given out.
A quantitative evaluation method using variogram function and singular value decomposition is proposed in this paper. Texture features of the images are extracted from calculating the variogram function value, and then a quantitative evaluation system is obtained using the texture features to weight sum of the singular value. Test samples are used in this paper to test the reliability of the system, and the accuracy is 93.3%,so the quantitative evaluation system can be used to assist doctors to evaluating the effect of the liver cancer radiofrequency treatment more accurately.
The region of interest of the ultrasound images we get are all manual segmentation. Image segmentation algorithm is studied in this paper and we hope that we can use automatic segmentation instead of manual segmentation to achieve automatic detection of cancer. Otsu threshold method, maximum entropy threshold method and fuzzy partition entropy method are used in ultrasound image segmentation and experiment results are given out.
引文
[1]唐安球.肝癌射频消融治疗现状[J].医学文选,2006,25(1)：156-158.
[2]倪雪君,周国雄,成建萍.射频消融治疗肝癌机制的研究进展[J].世界华人消化杂志,2008,16(30)：3416-3420.
[3]吴立顺,菅喜歧.超声图像诊断技术的发展及应用现状[J].医疗卫生装备,2007,28(12)：35-36.
[4]C. M. Wu, Y.C.Chen, K. Sheng. Texture Features for Classification of Ultrasonic Liver Images[J].IEEE Transactions on Medical Imaging.1992,11(2):141-152.
[5]田捷,包尚联,周明全.医学影像处理与分析[M].北京：电子工业出版社,2003.
[6]章毓晋.图像工程(上册)-图像处理与分析[M].北京：清华大学出版社,1998.
[7]秦襄培.MATLAB图像处理与界面编程[M].北京：电子工业出版社,2009.
[8]R. M. Haralick, K. Shanmugam, H. Dinstein. Textural Features for Image Classification[J]. IEEE Transaction on Systems,Man and Cybernetics.1973,3(6):610-621.
[9]冯建辉,杨玉静.基于灰度共生矩阵提取纹理特征图像的研究[J].北京测绘.2007,3：19-22.
[10]朱红娟,李芬华,赵真等.灰度行程统计法分析肝超声图像的研究[J].山西电子技术.2005,3：22-24.
[11]Michael Unser. Textural Classification and Segmentation Using Wavelet Frames[J]. IEEE Transaction on Image Processing.1995,4(11):1549-1560.
[12]S. Arivazhagan, L. Ganesan, V. Angayarkanni.Color Textural Classification Using Wavelet Transform[C].IEEE Proceedings of the Sixth International Conference on Computational Intelligence and Multimedia Applications.2005.
[13]C.C.Chen, D.C.Chen. Multi-resolutional gobor filter in texture analysis[J].Pattern Recognition Letters.1996,17:1069-1076.
[14]M. W. Jian, L. Liu, F. Guo. Texture Image Classification Using Perceptual Texture Feature and Gabor Wavelet Features[C].Asia-Pacific Conference on Information Processing.2009, 55-58.
[15]A. Ahmadian, A. Mostafa, M. D. Abolhassani.A method for texture classification of ultrasonic liver images based on Gabor Wavelet[J].ICSP'04 Proceedings.2004,971-974.
[16]Feng Gui,Lin Qi Wei.Application of variogram function in image analysis[J].ICSP'04 Proceedings.2004,1099-1102.
[17]Qi Chen, Peng Gong. Automatic Variogram Parameter Extraction for Textural Classification of the Panchromatic IKONS Imagery[J].IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING.2004,5(42)1106-1115.
[18]E. Sertel,S.Kaya, P. J. Curran. Use of Semivariograms to Identify Earthquake Damage in an Urban Area[J].IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING.2007,6(45) 1590-1594.
[19]A. Y. Boubas,M. Bettayeb.A Multicomponent Image Quality Assessment Based on Singular Value Decomposition[J].IEEE International Conference on Signal and Communications.2007,141-144.
[20]姚敏.数字图像处理[M].北京：机械工业出版社,2006.
[21]Shuhong Jiao, Xueguang Li,Xin Lu. An Improved Ostu Method for Image Segmentation[J].ICSP2006 Proceedings.2006.
[22]马英辉.基于改进Otsu法的自动阈值分割方法[J].科技咨询导报,2007,21：47-48.
[23]周德龙,潘泉,张洪才等.最大熵阈值处理算法[J].软件学报,2001,12(9)：1420-1422.
[24]Wang Feng-chao.An Improved 2-D Maximum Entropy Threshold Segmentation Method Based on PSO[J].IEEE.2009.
[25]C.H.Cheng. A Simple Fuzzy Group Decision Making Method[J].IEEE International Fuzzy Systems Conference Proceedings.1999,910-915.
[26]Ming Qing. Analysis and Design of a Fuzzy System Based on Fuzzy Entropy of Fuzzy Partitions[J].IEEE International Fuzzy Systems Conference,2001,1503-1506.
[2]倪雪君,周国雄,成建萍.射频消融治疗肝癌机制的研究进展[J].世界华人消化杂志,2008,16(30)：3416-3420.
[3]吴立顺,菅喜歧.超声图像诊断技术的发展及应用现状[J].医疗卫生装备,2007,28(12)：35-36.
[4]C. M. Wu, Y.C.Chen, K. Sheng. Texture Features for Classification of Ultrasonic Liver Images[J].IEEE Transactions on Medical Imaging.1992,11(2):141-152.
[5]田捷,包尚联,周明全.医学影像处理与分析[M].北京：电子工业出版社,2003.
[6]章毓晋.图像工程(上册)-图像处理与分析[M].北京：清华大学出版社,1998.
[7]秦襄培.MATLAB图像处理与界面编程[M].北京：电子工业出版社,2009.
[8]R. M. Haralick, K. Shanmugam, H. Dinstein. Textural Features for Image Classification[J]. IEEE Transaction on Systems,Man and Cybernetics.1973,3(6):610-621.
[9]冯建辉,杨玉静.基于灰度共生矩阵提取纹理特征图像的研究[J].北京测绘.2007,3：19-22.
[10]朱红娟,李芬华,赵真等.灰度行程统计法分析肝超声图像的研究[J].山西电子技术.2005,3：22-24.
[11]Michael Unser. Textural Classification and Segmentation Using Wavelet Frames[J]. IEEE Transaction on Image Processing.1995,4(11):1549-1560.
[12]S. Arivazhagan, L. Ganesan, V. Angayarkanni.Color Textural Classification Using Wavelet Transform[C].IEEE Proceedings of the Sixth International Conference on Computational Intelligence and Multimedia Applications.2005.
[13]C.C.Chen, D.C.Chen. Multi-resolutional gobor filter in texture analysis[J].Pattern Recognition Letters.1996,17:1069-1076.
[14]M. W. Jian, L. Liu, F. Guo. Texture Image Classification Using Perceptual Texture Feature and Gabor Wavelet Features[C].Asia-Pacific Conference on Information Processing.2009, 55-58.
[15]A. Ahmadian, A. Mostafa, M. D. Abolhassani.A method for texture classification of ultrasonic liver images based on Gabor Wavelet[J].ICSP'04 Proceedings.2004,971-974.
[16]Feng Gui,Lin Qi Wei.Application of variogram function in image analysis[J].ICSP'04 Proceedings.2004,1099-1102.
[17]Qi Chen, Peng Gong. Automatic Variogram Parameter Extraction for Textural Classification of the Panchromatic IKONS Imagery[J].IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING.2004,5(42)1106-1115.
[18]E. Sertel,S.Kaya, P. J. Curran. Use of Semivariograms to Identify Earthquake Damage in an Urban Area[J].IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING.2007,6(45) 1590-1594.
[19]A. Y. Boubas,M. Bettayeb.A Multicomponent Image Quality Assessment Based on Singular Value Decomposition[J].IEEE International Conference on Signal and Communications.2007,141-144.
[20]姚敏.数字图像处理[M].北京：机械工业出版社,2006.
[21]Shuhong Jiao, Xueguang Li,Xin Lu. An Improved Ostu Method for Image Segmentation[J].ICSP2006 Proceedings.2006.
[22]马英辉.基于改进Otsu法的自动阈值分割方法[J].科技咨询导报,2007,21：47-48.
[23]周德龙,潘泉,张洪才等.最大熵阈值处理算法[J].软件学报,2001,12(9)：1420-1422.
[24]Wang Feng-chao.An Improved 2-D Maximum Entropy Threshold Segmentation Method Based on PSO[J].IEEE.2009.
[25]C.H.Cheng. A Simple Fuzzy Group Decision Making Method[J].IEEE International Fuzzy Systems Conference Proceedings.1999,910-915.
[26]Ming Qing. Analysis and Design of a Fuzzy System Based on Fuzzy Entropy of Fuzzy Partitions[J].IEEE International Fuzzy Systems Conference,2001,1503-1506.