Joint Segmentation of MR and CT scans for Gynecologic Cancer Brachytherapy

描述(申请人提供)：将放射性同位素直接植入子宫颈癌或阴道癌以根除癌症，需要在肿瘤内插入中空涂抹器。虽然以前的普通X光胶片描绘了涂抹器的位置，并被用来计算输送到涂抹器周围区域的剂量，但三维CT在20世纪90年代给肿瘤可视化带来了革命性的变化。然而，MRI对妇科肿瘤的定义更清晰，尽管MRI在妇科近距离放射治疗中的应用还处于初级阶段。过去五年的研究表明，与MR相比，单独使用CT高估了癌症周围正常组织的范围。相比之下，MR允许放射肿瘤学家在应该接受最高辐射剂量的目标体积周围画出更紧密的边界，同时保留更大数量的周围正常组织。我们医院目前的做法是在手术室放置敷贴器后，获取3D图像来指导放射治疗计划。患有大肿瘤的患者既要接受CT检查，也要接受核磁共振检查。商业软件将MRI和CT相互对准，并用于勾勒出目标癌症和周围有风险的器官的轮廓。生成治疗计划，并使用该计划向患者提供辐射。当病人在麻醉下等待治疗时，目前的对准和轮廓方法需要放射肿瘤学家和物理学家提供大量的专家意见。该项目将使用如下先进的图像分析方法来提高这一工作流程的效率。1)我们将开发一台EM-ESP节段器，用于同时分割登记的MR和CT扫描，用于妇科近距离放射治疗计划。该分割器将构建和使用关于图像强度和关键解剖结构相对于近距离治疗器的几何形状的联合先验概率分布，这在图像中是清晰可见的。2)我们将对2004-2006年间由本院妇科放射肿瘤科主任阿基拉·维斯瓦纳坦博士使用MR引导的近距离放射治疗的25名患者的数据库进行回顾性应用EM-ESP节段仪。轮廓是在这些扫描上手动绘制的，我们将根据这些来测试我们的方法。该项目的结果将是一个算法，可以自动在MR和CT图像中勾画出宫颈、膀胱、直肠、乙状结肠和小肠，用于妇科近距离放射治疗。据我们所知，提出的算法将是第一个提供这一功能的算法。((！

DESCRIPTION (provided by applicant): The placement of radioactive isotopes directly into a cancer of the uterine cervix or vagina in order to eradicate the cancer, requires insertion of a hollow applicator into the tumor. Though previously plain x-ray film depicted the position of the applicator, and was used to calculate the dose to deliver to an area around the applicator, 3D CT revolutionized tumor visualization in the 1990s. However, MRI more clearly defines gynecologic tumors, though the application of MRI to gynecologic brachytherapy is in its infancy. Research in the last five years has shown that using CT alone overestimates the extent of normal tissue that surrounds the cancer compared to MR. In comparison MR allows the radiation oncologist to draw tighter boundaries around the target volume that should receive the highest radiation dose, while sparing larger amounts of surrounding normal tissue. Current practice at our hospital after applicator insertion in the operating room is to obtain a 3D image to guide radiation treatment planning. Patients with large tumors undergo both a CT and an MRI. Commercial software aligns the MRI and CT to each other, and is used to outline target cancer and surrounding organs at risk. A treatment plan is generated and radiation is delivered to the patient using this plan. Current alignment and contouring methods require significant expert input from the radiation oncologist and physicist while the patient waits for treatment under anesthesia. This project will improve the efficiency of this workflow using advanced image analysis methods as follows. 1) We will develop an EM-ESP segmenter to simultaneously segment registered MR and CT scans for gynecologic brachytherapy planning. This segmenter will construct and use joint prior probability distributions on image intensities and the geometry of key anatomic structures relative to the brachytherapy applicator, which is clearly visible in the images. 2) We will apply the EM-ESP segmenter retrospectively on a database of 25 patients who were treated during 2004-2006 using MR guided brachytherapy by Dr. Akila Viswanathan, Chief of Gynecologic Radiation Oncology at our hospital. Contours were drawn manually on these scans, and we will test our method against these. The result of this project will be an algorithm that can automatically delineate the cervix, the bladder, the rectum, the sigmoid colon, and the small bowel in MR and CT images for gynecologic brachytherapy. To the best of our knowledge, the proposed algorithm will be the first to provide this functionality. ( ( !