Robust BCT for Clinical Use

描述（由申请人提供）：骨质疏松症是50岁以上人口中超过50%的主要公共卫生威胁。尽管骨质疏松症很重要，但它在很大程度上没有得到治疗，只有不到20%的推荐检测者得到了筛查。由于联邦医疗保险对双能x线骨密度测量（DXA，目前的临床标准）的报销大幅削减，DXA对骨折预测的敏感性低于50%，并且随着美国老龄化人口规模的迅速增加，迫切需要比DXA更敏感的方法来评估骨折风险。生物力学计算机断层扫描（BCT）已成为DXA的强大替代品。这种基于CT的技术根据患者的CT扫描创建了骨骼的结构“有限元”模型，并将该模型与虚拟力相结合，以提供骨骼强度的估计。BCT在尸体研究中得到了很好的验证，并且比DXA的骨矿物质密度更能预测骨强度，在临床研究中也被证明对骨质疏松性骨折有很高的预测作用。然而，稳健性仍然是一个问题——该技术是否可以被非专家在研究和临床环境中轻松使用？为了解决这个问题，本研究的总体目标是提高我们软件的鲁棒性，这样它就可以自动分析来自各种CT扫描仪的扫描，并使用各种CT采集方案，包括限制患者辐射暴露的新低剂量方案。这样一个强大的BCT诊断工具可以作为许多其他目的的CT检查（如CT结肠镜检查、骨盆检查、腹部检查和脊柱检查）的补充“附加”分析，从而降低医院成本，不增加对患者的辐射，不需要改变CT采集方案，因此大大增加了可以低成本筛查的患者数量。具体来说，我们建议在这个第一阶段项目中结合计算机视觉、CT扫描和生物力学方面的专业知识，以开发一种自动的“无影”交叉校准CT扫描的方法，用于稳健的椎体强度评估。专注于脊柱，我们的主要任务是执行一系列临床研究，在这些研究中，患者使用各种CT采集方案进行两次扫描；开发一个定制的外部校准模型，并使用它来确定各种CT采集参数对扫描标准化的影响；并使用机器学习技术开发脊柱的“统计地图集”，以实现所有图像处理的自动化。我们将结合这些努力开发一种无影BCT方法，该方法可以解释由于CT扫描仪和采集协议（包括低剂量协议）的变化而导致的图像质量差异，并且以高度自动化的方式完成，需要最少的用户专业知识和输入。如果该项目成功，未来的工作将进一步完善技术，将其扩展到髋关节和肌肉和其他软组织的定量分析，并解决临床监测纵向变化的稳健性。公共卫生相关性：髋部骨折后一年的死亡率高达30%，每年的经济负担超过170亿美元，骨质疏松性骨折是一种使人衰弱的疾病，对我们老龄化社会的影响越来越大。早期识别骨折风险可以指导预防和治疗，BCT将为这种检测提供一种方法，其灵敏度和特异性是基于DXA的骨密度测量所缺乏的。然而，CT的较大辐射暴露限制了这种诊断的市场。拟议的项目将产生一种强大的诊断测试，大大降低对患者的辐射剂量，并且在某些实施中，通过使用已经为其他医疗目的订购的CT扫描，完全消除了额外的辐射。该产品的成功开发将扩大受益于更准确和敏感的骨折风险预测的人群，扩大O. N. Diagnostics的业务市场，并在骨质疏松症的预防保健和治疗方面取得重要进展。

DESCRIPTION (provided by applicant): Osteoporosis is a major public health threat for over 50% of the population over age 50. Despite its importance, osteoporosis is largely under-treated, with less than 20% of those recommended for testing being screened. With substantial reimbursement cuts being introduced by Medicare for bone densitometry by dual energy X-ray absorptiometry (DXA, the current clinical standard), with a sensitivity of DXA for fracture prediction of less than 50%, and with the rapidly increasing size of the aging population of the U.S., there is an urgent need for additional and more sensitive modalities than DXA for clinical assessment of fracture risk. Biomechanical Computed Tomography (BCT) has emerged as a powerful alternative to DXA. This CT-based technology creates a structural "finite element" model of a patient's bone from their CT scans, and subjects that model to virtual forces in order to provide an estimate of the strength of the bone. Well validated in cadaver studies and being a better predictor of bone strength than is bone mineral density by DXA, BCT has also been shown to be highly predictive of osteoporotic fractures in clinical research studies. However, robustness remains an issue - can the technique be used easily by non-experts in research and clinical environments? Addressing this issue, the overall goal of this research is to improve the robustness of our software, such that it can automatically analyze scans from a wide range of CT scanners and using a wide variety of CT acquisition protocols, including new low-dose protocols that limit radiation exposure to the patient. Such a robust BCT diagnostic tool could then be offered as a supplementary "add-on" analysis to many types of CT exams taken for other purposes such as CT colonography, pelvic, abdominal, and spine exams, thus reducing hospital costs, incurring no addition radiation to the patient, requiring no change in the CT acquisition protocols, and therefore greatly increasing the number of patients that could be screened at low cost. Specifically, we propose in this Phase-I project to combine expertise in computer vision, CT scanning, and biomechanics in order to develop an automated method of "phantomless" cross-calibration of CT scans for robust vertebral strength assessment. Focusing on the spine, our major tasks are to perform a series of clinical studies in which patients are scanned twice using a variety of CT acquisition protocols; develop a custom external-calibration phantom and use that to determine the effects of various CT acquisition parameters on scanning standardization; and use machine learning techniques to develop a "statistical atlas" of the spine for automation of all image processing. We will combine these efforts to develop a phantomless BCT method that accounts for differences in image quality due to variations in CT scanners and acquisition protocols, including low-dose protocols, and that does so in a highly automated fashion requiring minimal user expertise and input. Should this project be successful, future work will further refine the techniques, extend them to the hip and quantitative analysis of muscle and other soft tissues, and address robustness of longitudinal changes for clinical monitoring. PUBLIC HEALTH RELEVANCE: With a mortality rate up to 30% one year after hip fracture, and an economic burden exceeding $17 billion annually, osteoporotic fracture is a debilitating condition whose impact on our aging society is growing. Early identification of those at risk for fracture can guide prevention and treatment, and BCT will provide a means for such detection with a sensitivity and specificity lacking in DXA based bone densitometry. The greater radiation exposure from CT, however, limits the market for such a diagnostic. The proposed project will result in a robust diagnostic test that significantly lowers radiation dose to the patient, and in some implementations, completely eliminates additional radiation by using CT scans already ordered for other medical purposes. Successful development of this product will broaden the pool of individuals who will benefit from a more accurate and sensitive fracture risk prediction, expand the market for O. N. Diagnostics' business, and result in an important advance in the preventative care and treatment of osteoporosis.