摘要
目的:探讨BP神经网络算法在新疆高发病哈萨克族食管癌X射线图像纹理特征和形状特征分型中的应用。方法:选取哈萨克族正常食管图像、溃疡型食管癌图像和髓质型食管癌图像各100张,利用灰度共生矩阵算法和形状不变距算法分别提取图像的纹理和形状特征。然后,使用BP神经网络算法构造一个分类器,对正常食管和两种中晚期食管癌图像进行分类研究。结果:共提取了14维哈萨克族食管癌X射线图像纹理和形状特征向量,应用BP神经网络算法进行哈萨克族食管癌X射线图像分类实验,基于灰度共生矩阵算法的纹理特征分类准确率为85.333%,基于Hu不变距算法的形状特征分类准确率为65.333%,而纹理和形状综合特征的分类准确率达到了97.667%。结论:本研究提取基于灰度共生矩阵算法和Hu不变距算法的食管癌图像纹理和形状特征,通过构造BP神经网络分类器对食管癌医学图像进行分型研究。结果表明BP神经网络对综合特征的分类准确率较高,为临床医生诊断食管癌提供了参考,也为后期研发食管癌医学图像计算机辅助诊断系统奠定了基础。
Object:This paper explore the application of BP neural network algorithm combined with image texture and shape feature extraction in X ray image classification of Kazak esophageal cancer in Xinjiang. Methods:The images of normal esophagus, ulcerative carcinoma of esophagus and medullary esophagus cancer were selected, each of which was 100, then the image texture and shape features are extracted by gray co-occurrence matrix method and shape invariant distance method respectively. After that,using BP neural network algorithm to construct a classifier was used to classify the normal esophagus and two kinds of advanced esophageal carcinoma. Results:A total of 14 dimensional features are extracted, using a single feature algorithm for classification, gray level co-occurrence matrix classification accuracy is85.333%, Hu invariant distance classification accuracy is 65.333%. The comprehensive feature classification accuracy was 97.667%,which is more suitable for classification of normal esophagus and middle and advanced esophageal carcinoma. Conclusion:In this study, the image texture and shape feature extraction algorithm is combined with neural network algorithm to analyze characteristics and typing of normal esophagus, ulcerative and medullary esophageal carcinoma. This method has higher classification accuracy, which lays a foundation for the development of computer aided diagnosis system for esophageal cancer medical images.
引文
[1] 曹小琴,孙喜斌.食管癌发病水平及变化趋势[J].中国肿瘤临床,2016,43(21):932-936.
[2] 王志鹏,黄德静,钟武,等.食管癌与2型糖尿病相关性的研究进展[J].新疆医学,2017,47(2):119-122.
[3] 杨爱华.食管癌的临床分析和治疗探究[J].中国伤残医学,2014,137(14):168-169.
[4] 孔喜梅,木拉提·哈米提,严传波,等.基于PCA和C4.5决策树的新疆哈萨克族食管癌图像鉴别研究[J].科技通报,2016,32(9):52-57.
[5] 刘京华.基于贝叶斯模型的医学影像分类技术的研究与应用[D].电子科技大学,2016.
[6] 黄雪梅,唐治德,赵一凡,等.BP网络研究及其在肺癌诊断系统中的应用[J].重庆大学学报,2005,28(1):42-44.
[7] 任国贞.基于灰度共生矩阵的纹理提取及分类研究[D].山东科技大学,2015.
[8] 徐胤,乐琚,袁浩巍. 基于不变矩特征和RBF神经网络的图像识别[J].上海电气技术,2015,8(3):33-36.