论文部分内容阅读
目的为了提高分类和检索准确率从哈萨克族食管癌X射线医学图像中提取感兴趣区域。方法通过40幅新疆哈萨克族食管癌X射线医学图像中得出新疆哈萨克族食管癌医学图像直方图特征并利用此特征分别采用区域增长法和全阈值法分割100新疆哈萨克族食管癌医学图像,并利用面积大小差异和平均表面距离评价分割结果。结果分割后的图像与手工分割的图像进行比较评价得出区域增长法与手工分割图像面积平均相差4.1606%,平均表面距离相差1个像素点而全阈值法分割后图像面积平均相差13.056%,平均表面距离相差3个像素点。结论区域增长法比较适合分割新疆哈萨克族食管癌X射线医学图像。此研究不仅能提高新疆哈萨克族食管癌辅助诊断系统的诊断准确率并对以后的食管癌医学图像分割研究奠定了基础。
Objective To improve the accuracy of classification and retrieval, extract the region of interest from X-ray medical images of Kazakh esophageal cancer. Methods Forty Xinjiang Kazak esophageal cancer X-ray medical images were obtained from Xinjiang Kazakh medical image histogram features and using this method, the regional growth method and the full threshold method were used to segment 100 Xinjiang Kazakh esophageal cancer medical images The segmentation results were evaluated using the difference in area size and the average surface distance. Results Compared with manual segmentation, the results of segmentation and image segmentation showed that the difference between regional growth method and manual segmentation was 4.1606%, the average surface distance was 1 pixel, while the average threshold area was 13.056% The surface distance difference of 3 pixels. Conclusion Regional growth method is more suitable for the segmentation of Xinjiang Kazakh esophageal cancer X-ray medical images. This study not only can improve the diagnostic accuracy of Kazakh-assisted diagnosis system of esophageal carcinoma in Xinjiang, but also lay the foundation for the future study of medical image segmentation of esophageal cancer.