对膀胱肿瘤浸润深度进行无创检测,为膀胱肿瘤的检测及分期判断提供参考.基于膀胱MRI图像,利用尿液和膀胱壁的天然密度对比,结合统计分析,得到在肿瘤及膀胱壁组织有统计差异的纹理特征.通过提取肿瘤及周围膀胱壁组织的均值、均匀度、标准差、粗糙度、自协方差系数和对比度等特征,再经分类器判断,得到感兴趣区域肿瘤浸润情况的伪彩图.就16位患者MRI扫描数据及术后病理分析结果,用于测试的56幅图像中,判断准确率为82.1%;15位患者的标记结果与病理分期相符,正确率为93.8%,与病理分析吻合较好.利用分类器和纹理特征判别肿瘤区域的属性,实现了膀胱肿瘤浸润深度标记和边界的初步划分,为膀胱肿瘤的无创检测提供了新的可能手段.%A MRI-based detection scheme was proposed for the detection of bladder tumor and its invasion depth,which has a potential for noninvasive detection of bladder tumor and its stage.Since MRI bladder images could provide natural contrast between urine and bladder wall, there were some texture features demonstrating statistically significant difference between tumor tissue and wall tissue.Among them, mean, standard deviation,uniformity,covariance and contrast, were selected for classification.Then, a labeling scheme was proposed for obtaining the pseudo-color mapping of bladder tumor invasion.Experiments based on MRI datasets of 16 patients were carried out to evaluate the scheme.It was revealed that among 56 testing images, classification and labeling results of 46 images (82.1％ ) were consistent with their histopathological results from postoperative biopsy,while labeling results of 15 out of 16 patients consistented well with histopathological tests,with an accuracy of 93.8％.With texture-based classification, the proposed method implemented preliminary detection of bladder tumor and its invasion depth,which might be a novel and promising tool for noninvasive tumor detection with virtual cystoscopy.
展开▼
