Optimizing automated MRI measures of atrophy in neurodegenerative disease

描述（由申请人提供）：阿尔茨海默病在美国折磨着大约510万人，随着预期寿命的增加，预计这一数字将继续上升（医学研究所，2008）。除了患者及其护理人员（通常是家庭成员）的个人费用外，到2050年，护理痴呆症患者的费用预计将从2011年的1870亿美元增加到1.1万亿美元。目前最广泛接受的自动化方法测量的海马在阿尔茨海默病是FreeSurfer。在Brain Image Analysis，LLC，我们使用我们的自动化管道BRAINS AutoWorkup处理了来自ADNI 1数据集（>3500次扫描）的MRI数据。在我们的分析中发现，与标准FreeSurfer和FreeSurfer中的纵向流相比，我们的方法检测到更高的年萎缩率，相对标准差更低。这导致临床试验每组所需受试者的估计值大幅减少，从使用FreeSurfer（纵向和横截面工作流程）的131或204例减少到使用BRAINS ANN方法的56例。本申请描述了我们打算如何实施和测试纵向人工神经网络方法，并测试我们目前的海马分割对阿尔茨海默病研究人员和海马专家的合作准备。这将为Brain Image Analysis，LLC提供所需的信息，以表明我们的皮层下分割方法在阿尔茨海默病及其治疗研究中的商业图像处理领域是最可行的。该项目的第二阶段将包括将这些方法实施到最新版本的BRAINS和额外的程序基础设施中，以使我们的方法在更大的商业规模上可用，以及在大型数据集上实施以探索临床相关性。在这个项目上与我们合作的将是著名的神经科学研究人员。其中包括Brains开发团队的长期领导者Vincent Magnotta博士和Nancy Andreasen博士，以及在支持临床试验的大型成像研究方面具有丰富经验的统计学家Doug Langbehn博士。 公共卫生相关性：在美国，阿尔茨海默病折磨着估计510万人，并且随着预期寿命的增加，预计这个数字将继续上升（Hebert等人，2003年）。除了患者及其护理人员（通常是家庭成员）的个人费用外，到2050年，护理痴呆症患者的费用预计将从2011年的1870亿美元增加到1.1万亿美元（医学研究所，2008年）。阿尔茨海默病神经影像学倡议是一项正在进行的纵向多中心研究，旨在开发临床，成像，遗传和生化生物标志物，用于早期检测和跟踪阿尔茨海默病。与ADNI数据集相关的工作的主要焦点之一是通过提供敏感的生物标志物来增强临床试验，例如萎缩率的海马测量，可以在临床试验中监测疾病改善效果。在Brain Image Analysis，LLC，我们使用我们目前的方法来分析ADNI MRI数据集，并发现我们的方法对海马萎缩非常敏感，与FreeSurfer相比，需要不到1/2的受试者数量才能在临床试验中检测到萎缩率降低25%。在本申请中，我们建议进一步开发和加强我们的方法，作为监测疾病进程的商业服务，并作为旨在改变疾病进程的药物的临床试验中的生物标志物。这个第一阶段SBIR应用程序提出了以下目标，以更全面地评估我们目前的方法，将其扩展到纵向管道，并根据阿尔茨海默氏症研究社区正在开发的“共识”海马定义对其进行评估。目的1 -在BRAINS中实现纵向人工神经网络海马分割，并评估其在阿尔茨海默病海马萎缩中的准确性和敏感性，与用于处理初步数据的标准ANN进行比较。AIM 2-将我们目前的海马分割结果与Giovanni Frisoni和Clifford Jack开发的“共识”方法进行比较。目的3-计算目的1和2中开发的方法检测纵向萎缩率变化的能力，并与FreeSurfer和其他可用方法进行比较。在这个项目上与我们合作的将是著名的神经科学研究人员。其中包括Brains开发团队的长期领导者Vincent Magnotta博士和Nancy Andreasen博士，以及在支持临床试验的大型成像研究方面具有丰富经验的统计学家Doug Langbehn博士。如果第一阶段成功，该项目的第二阶段将包括将这些方法应用到最新版本的BRAINS和额外的程序基础设施中，以使我们的方法在更大的商业规模上可用，以及在大型ADNI数据集上的应用，以探索与监测疾病进展和临床试验相关的临床相关性。

DESCRIPTION (provided by applicant): Alzheimer's disease afflicts an estimated 5.1 million people in the United States, and as life expectancies increase it is anticipated that this number will continue to rise (Institute of Medicine, 2008). Besides the personal toll on the patients and their caregivers (who often are family members), the cost of caring for those with dementia is anticipated to grow from the 2011 estimate of $187 billion to $1.1 trillion by the year 2050. Currently the most widely accepted automated method for measurement of the hippocampus in Alzheimer's disease is FreeSurfer. At Brain Image Analysis, LLC we have processed the MRI data from the ADNI 1 dataset (>3500 scans) using our automated pipeline, BRAINS AutoWorkup. In our analysis have found that, in comparisons with both the standard FreeSurfer and the longitudinal stream in FreeSurfer, our methods detect a higher annual atrophy rate with a lower relative standard deviation. This leads to a substantial reduction in the estimates of subjects needed per arm of a clinical trial from 131 or 204 using FreeSurfer (longitudinal and cross-sectional workflows) to 56 using BRAINS ANN methods. This application describes how we intend to implement and test a longitudinal ANN method and test our current hippocampal segmentations against those being prepared by a collaboration of Alzheimer's disease researchers and hippocampus experts. This will provide Brain Image Analysis, LLC with the information it needs to show our methods of subcortical segmentation are the most feasible in the field for commercial image processing in the study of Alzheimer's disease and its treatment. Phase II of this project will encompass implementation of these methods into the newest version of BRAINS and the additional program infrastructure to make our methods available on a larger commercial scale, as well as implementation on a large dataset to explore clinical correlates. Working with us on this project will be renowned neuroscience researchers. These include Dr. Vincent Magnotta and Dr. Nancy Andreasen, long-term leaders of development team for BRAINS, and Dr. Doug Langbehn, serving as our statistician with substantial experience in large imaging studies designed to support clinical trials. PUBLIC HEALTH RELEVANCE: Alzheimer's disease afflicts an estimated 5.1 million people in the United States, and as life expectancies increase it is anticipated that this number will continue to rise (Hebert et al., 2003). Besides the personal toll on the patients and their caregivers (who often are family members), the cost of caring for those with dementia is anticipated to grow from the 2011 estimate of $187 billion to $1.1 trillion by the year 2050 (Institute of Medicine, 2008). The Alzheimer's Disease Neuroimaging Initiative is an ongoing, longitudinal, multicenter study designed to develop clinical, imaging, genetic and biochemical biomarkers for the early detection and tracking of Alzheimer's disease. One of the main focuses of work related to the ADNI dataset is to enhance clinical trials through providing sensitive biomarkers, such as hippocampus measures of atrophy rates, with which disease modifying effects may be monitored in clinical trials. At Brain Image Analysis, LLC we have used our current methods to analyze the ADNI MRI dataset and have found our methods to be very sensitive to hippocampal atrophy, requiring less than 1/2 the number of subjects compared to FreeSurfer to detect a 25 percent reduction in atrophy rates in a clinical trial. In this applicatin we propose to further develop and strengthen our methods as commercially available service in monitoring disease course and as biomarkers in clinical trials of agents designed to alter the course of the disease. This Phase I SBIR application proposes the following aims to more completely assess our current methods, expand them into a longitudinal pipeline, and evaluate them against a "consensus" hippocampal definition being developed in the Alzheimer's research community. AIM 1 - Implement a longitudinal artificial neural network hippocampal segmentation in BRAINS and assess its accuracy and sensitivity in following hippocampal atrophy in Alzheimer's disease, in comparison that the standard ANN used to process the preliminary data. AIM2 - Compare the results of our current hippocampal segmentation with the "consensus" methods being developed by Giovanni Frisoni and Clifford Jack. AIM3 - Calculate the ability to detect changes in longitudinal atrophy rates for the methods developed in Aims 1 and 2, and compare with those of FreeSurfer and other available methods. Working with us on this project will be renowned neuroscience researchers. These include Dr. Vincent Magnotta and Dr. Nancy Andreasen, long-term leaders of development team for BRAINS, and Dr. Doug Langbehn, serving as our statistician with substantial experience in large imaging studies designed to support clinical trials. Should Phase I be successful, Phase II of this project will encompass implementation of these methods into the newest version of BRAINS and the additional program infrastructure to make our methods available on a larger commercial scale, as well as application on the large ADNI dataset to explore clinical correlates relevant to monitoring disease progression and clinical trials.