基于胃镜图像的食管早癌病灶计算机辅助诊断方法研究

发布时间：2018-11-09 20:32

【摘要】：要改善食管癌的治疗效果,最有效的途径就是提高常规胃镜检查的普及程度,这意味着消化内镜医师将要承担更加繁重的诊断任务。计算机辅助诊断方法能减轻医生的工作负担,减少人为因素可能导致的失误。本文利用图像处理、模式识别的方法实现了对普通电子胃镜图像中食管早癌病灶的检测,提高常规胃镜检查的效率和准确性。首先,考虑到主流的图像处理算法不适用于胃镜图像中区域提取,本文基于病变区域内像素间变化大的特点设计了一种简单、快速的感兴趣区域(ROI)提取方法;另外,本文开发了新的特征算子来描述每个感兴趣区域的整体和局部特点;最后,本文为了尽可能地降低漏诊率,对Adaboost算法中原本的样本权值初始化方案进行了修改,使得强分类器的训练过程中更加注重敏感性。实验结果表明,本文的方法检测食管早期癌症的敏感性和特异性分别为93.6%和77.2%,相对其他方法具有明显的优势。此外,本文方法计算简单、快速,更加符合临床应用方面的实时性要求。
[Abstract]:To improve the therapeutic effect of esophageal cancer, the most effective way is to increase the popularity of routine gastroscopy, which means that endoscopists will take on more heavy diagnostic tasks. The computer-aided diagnosis method can reduce the burden of doctors' work and the mistakes that may be caused by human factors. Image processing and pattern recognition are used to detect early esophageal cancer lesions in general electronic gastroscopy images, and the efficiency and accuracy of conventional gastroscopy are improved. Firstly, considering that the mainstream image processing algorithm is not suitable for the region extraction of gastroscope image, this paper designs a simple and fast (ROI) extraction method for the region of interest based on the large variation of pixels in the lesion region. In addition, a new feature operator is developed to describe the global and local characteristics of each region of interest. Finally, in order to reduce the missed diagnosis rate as much as possible, the original initialization scheme of sample weight in Adaboost algorithm is modified, which makes the training of strong classifier more sensitive. The results showed that the sensitivity and specificity of our method for detecting early esophageal cancer were 93.6% and 77.2% respectively. In addition, this method is simple, fast and more in line with the real-time requirements of clinical application.
【学位授予单位】：电子科技大学
【学位级别】：硕士
【学位授予年份】：2015
【分类号】：R735.1;TP391.41

【相似文献】