Computer-Aided Detection of Urinary Tract Cancer on MDCT Urography

描述(申请人提供)：尿路肿瘤是一种常见的癌症类型，可在患者中导致相当大的发病率和死亡率。在美国，膀胱癌和上尿道癌每年导致14800人死亡。预计2008年将诊断出71100例新的膀胱癌和上尿路癌病例。多排螺旋CT(MDCT)尿路造影术是目前非常有前景的早期发现膀胱癌和上尿路癌的成像方法，这可能是导致血尿的原因之一。血尿在老年患者中的患病率可高达19%。对放射科医生来说，解释通常超过400层的MDCT尿路造影(CTU)是一项艰巨的任务，他们必须视觉跟踪上、下尿路并寻找通常较小的病变。此外，一些膀胱病变可以在充满对比剂的膀胱区，而一些在无对比剂的区域。该项目的长期目标是开发一种有效的计算机辅助诊断(CADx)系统，以帮助放射科医生解释CTUS。在这个拟议的项目中，我们将专注于开发第一个计算机辅助检测(CAD)系统，用于检测CTU图像上的膀胱和上尿路病变。我们推测，使用CAD系统可以提高放射科医生在CTUS上诊断膀胱癌和上尿路癌的准确性。为了验证这一假设，我们将执行下列具体任务：(1)收集膀胱及上尿路良恶性病变的数据库；(2)开发新的计算机视觉技术来处理三维(3D)容积CTU；(3)开发检测膀胱病变的算法；(4)开发检测上尿路病变的算法；以及(5)比较观察者ROC研究在有和没有CAD的情况下CTU对膀胱和上尿路病变的检测准确性。为了完成这些任务，我们将开发新的图像分析技术，用于自动跟踪输尿管，并分割膀胱和输尿管的内壁和外壁。将专门设计新的方法来检测膀胱和输尿管中的候选病变。我们将设计方法和3D测量方法来估计膀胱壁厚度的不对称性和输尿管壁增厚的检测。将开发特征提取技术和稳健的分类方法，以利用提取的特征识别真阳性和消除假阳性病变。如果开发成功，该CAD系统可能会改善放射科医生在检测尿路上皮肿瘤以及解释血尿患者的CTU方面的表现，从而能够在早期发现更多的癌症。早期发现可改善患者的预后和生存率。

DESCRIPTION (provided by applicant): Urinary tract neoplasm is a common type of cancer that can cause substantial morbidity and mortality among patients. Bladder and upper urinary tract cancer causes 14800 deaths per year in the United States. It is expected that 71100 new bladder and upper urinary tract cancer cases would be diagnosed in 2008. Multi-detector row CT (MDCT) urography is currently a very promising imaging modality for early detection of bladder and upper urinary tract cancer, which can be a cause of hematuria. The prevalence of hematuria can be as high as 19% in elderly patients. Interpretation of MDCT urograms (CTU) that commonly exceeds 400 slices is a demanding task for radiologists who have to visually track the upper and lower urinary tract and look for lesions which usually are small in size. In addition, some bladder lesions can be in the bladder area filled with contrast and some in the area without contrast. The long term goal of the project is to develop an effective computer-aided diagnosis (CADx) system to assist radiologists in interpretation of CTUs. In this proposed project, we will concentrate on the development of the first computer-aided detection (CAD) system for the detection of bladder and upper urinary tract lesions on CTU images. We hypothesize that the use of CAD system can improve the radiologists' accuracy in detecting bladder and upper urinary tract cancer on CTUs. To test this hypothesis, we will perform the following specific tasks: (1) collect a database of bladder and upper urinary tract malignant and benign lesions; (2) develop new computer vision techniques to process 3-dimensional (3D) volumetric CTUs; (3) develop algorithms to detect bladder lesions; (4) develop algorithms to detect upper urinary tract lesions; and (5) compare the detection accuracy of bladder and upper urinary tract lesions on CTUs with and without CAD by observer ROC studies. In order to accomplish these tasks, we will develop new image analysis techniques for automated tracking of the ureter and segmentation of the inner and outer walls of the bladder and the ureter. New methods will be designed specifically for detection of lesion candidates in the bladder and the ureter. We will design methods and 3D measures for estimating asymmetries of the bladder wall thickness and detection of ureteral wall thickening. Feature extraction techniques and robust classification methods will be developed for identification of true positive and elimination of false positive lesions using the extracted features. If successfully developed, the CAD system can potentially improve the performance of the radiologists in detecting urothelial neoplasm as well as in interpreting CTU for patients with hematuria, allowing the detection of additional cancers at earlier stage. Early detection can improve the prognosis and survival of the patients.