基于特征融合的肝包虫病CT图像识别

CT image recognition of hepatic hydatid disease based on feature fusion

目的探讨特征融合方法在肝包虫病CT图像分类识别中的应用,旨在提高肝包虫病的诊断准确率。方法选取正常肝脏和单囊型肝包虫病CT图像各150张,对每幅图像采取空域与频域滤波算法、数学形态学算法和点处理,分别得到10幅特征子图像并对它们进行特征融合。对融合后的图像提取灰度和纹理特征,通过统计学分析筛选关键特征。结果对提取的10维特征进行统计学分析,得到正常肝脏和单囊型肝包虫CT融合图像之间完全没有交集的4个灰度和1个纹理特征取值范围,以此来区分肝包虫病与正常肝脏CT图像。结论从原始图像中提取特征子图像并进行融合,再对融合后图像提取特征的方法能够很好地区分识别正常肝脏和单囊型肝包虫病CT图像,为肝包虫病的早期诊断提供依据。

Objective To discuss the application of feature fusion method in the CT image classification and recognition of hepatic hydatid disease, and to improve the diagnostic accuracy of hepatic hydatid disease. Methods CT medical images of normal liver and single-cystic hepatic hydatid disease were selected,each 150 spatial and frequency-domain filtering algorithms, mathematical morphology algorithms and point processing were used to each image to obtain 10 feature images respectively and fused them effectively. Grayscale and texture features were extracted from each fused images, and the key features were selected by statistical analysis. Results Statistical analysis was performed on the extracted 10-dimensional features and obtained 4 gray scales and 1 texture features range with no intersection between the normal liver image and single-cystic hepatic hydatid fused image, thereby distinguishing the CT images of hepatic hydatid disease and normal liver. Conclusions Extracting feature sub-images from the original image and fusing them, extracting gray and texture features from the fused image can distinguish the CT images of normal liver and single-cystic hepatic hydatid, it also supplies evidence for the early diagnosis of hepatic hydatid disease.