Bladder Tumor Detection And Preliminary Staging Via MR Virtual Cystoscopy

Bladder tumor is one of the most common tumors of urinary system. According to the most recent statistics from Cancer Facts & Figures 2014, bladder cancer has become the fourth common cancer and the eighth leading cause of cancer death among American males. In China, bladder tumor becomes the 8th most common malignancy in men, and its incidence is climbing year by year. Moreover, bladder tumor has the characteristic of high recurrence rate after resection. Literatures show that the non-muscle invasive bladder tumor patients after surgical resection, approximately 50% of patients will find bladder tumor again in the next 18 months. As the gold standard for evaluating bladder tumors in clinics, optical cystoscopy(OC) may cause discomfort and iatrogenic injuries. Therefore, it is essential to find a noninvasive and convenient way for early detection of bladder cancer and follow-up management of tumor recurrence.Recently, with the fast development of medical imaging technologies, virtual cystoscopy(VCy) has shown its potential in detecting tumors. Several clinical trials have demonstrated that there is no significant difference of detection accuracy between VCy and OC when tumor size larger than 5mm.The CT and MRI are the common image modality for VCy. Compared to CT imaging, MRI-based VCy exhibits more advantages, where urine can be used as endogenous contrast medium and the scan is completely noninvasive and radioactive-free, and has attracted more attention recently.Computer-aided detection(CAD) based on image features of the bladder wall can further assist radiologists in improving accuracy and efficiency of abnormality detection. In previous studies, morphological features extracted from bladder wall, such as bladder wall thickness(BWT), curvedness(CV), shape index(SI) and bent rate(BR) have been used to detect wall abnormalities. In this study, we also used morphological features for tumor detection, and proposed a computer aided diagnose method for tumor staging.In this article, a bladder tumor detection and extraction pipeline has been established. Through this pipeline, bladder tumors could be detected effectively, and the tumor could be extracted from surrounding wall tissue which could be used to estimate tumor staging. The main work of the proposed pipeline can be summarized as follows:(1) Tumor detection based on bladder wall thicknessA novel FCM based method was used to detect tumors by using bladder wall thickness. Digital phantoms were used to set parameters of this method and test the effective of this method. Both on the phantoms and real patient datasets, this method showed its effectiveness.(2) Tumor detection based on multiple morphological featuresA multiple morphological features detection method was proposed in this article. A novel morphological feature, defined by the bent rate difference between two wall surfaces(named DBR), is first proposed. It can reflect curving similarity of two wall surfaces and has shown satisfying performance in detecting locally thickened regions. The proposed new feature, combined with previously used morphological features(SI, BWT, and BR), have been used for initial candidates detection. Preliminary results using MR datasets of patients validated its superior performance on sensitivity improvement and false positive reduction.(3) Extraction of tumor region and preliminary stagingConsidering the intensity difference between carcinomatous tissues and wall tissues in T2-weighted MR images, the Fuzzy C-Means segmentation with spatial information(s FCM)is used to further extract tumors from surrounding bladder wall tissues in detected regions, and the tumor staging were estimated.