Radiomics signatures and patient outcomes in intracerebral hemorrhage

项目摘要/摘要 下面的K23提案是为神经放射学家兼放射学助理教授Sam Payabvash博士准备的 在耶鲁大学。Payabvash博士是一名内科科学家，在 神经科学、神经成像和计算机视觉。他的职业目标是寻找新的治疗目标，并提供 脑血管疾病患者的个性化护理。脑出血(ICH)是最常见的 在没有有效治疗的情况下，毁灭性的脑血管疾病。到目前为止，脑出血风险的成像标志物- 分层和结果预测本质上是主观和描述性的，为 自动评估嵌入在医学图像中的成像缺陷。Payabvash博士的初步结果 已经论证了应用自动特征提取管道和机器的研究计划的可行性 利用医学图像信息进行早期风险分层和识别的学习算法 脑出血的潜在治疗靶点。在这项提案中，Payabvash博士将使用详细的临床和成像数据 3991名患者来自NIH资助的临床试验、在线档案和耶鲁大学、塔夫茨大学和 伦敦大学学院。他将应用机器学习算法来识别这些成像特征 基线头颅CT扫描中的脑出血与症状严重程度相关(目标1)。然后, 他将使用出血的成像特征来识别那些有早期扩张风险的患者 血肿(目标2a)或周围水肿(目标2b)。这两个“可修改”的不良结果指标是 考虑脑出血患者的潜在治疗目标。最后，他将结合入院临床信息和 建立长期结果预测的风险分层工具的成像特征(目标3)。在专家的带领下 凯文·谢斯博士(神经危重护理主任)、托德·康斯特布尔博士(核磁共振研究主任)的指导，以及 罗纳德·科夫曼博士(数学教授)，这个K23奖项将允许Payabvash博士(1)识别和 利用创新的神经游戏工具解决脑血管疾病中最紧迫的问题；(2)获得 在脑部扫描的高级统计分析方面的专业知识；以及(3)扩大他在机器学习和 用于医学图像评估的计算机视觉。Payabvash博士将接受神经成像方面的教学培训 统计分析、机器学习、深度神经网络和计算机视觉。拟议的研究和 职业发展计划利用耶鲁大学丰富的资源，包括区域协调 NIH StrokeNet中心、研究计算中心、高性能计算服务中心和 尖端的图像处理和分析基础设施。在颁奖期结束时，佩亚布瓦什博士 将处于有利地位，成为一名独立资助的调查员，在 旨在改善脑血管病人护理的先进神经成像技术和分析 疾病。

PROJECT SUMMARY / ABSTRACT The following K23 proposal is for Dr. Sam Payabvash, a Neuroradiologist and Assistant Professor of Radiology at Yale University. Dr. Payabvash is a physician-scientist with specialized expertise at the intersection of neuroscience, neuroimaging, and computer vision. His career goal is to find new treatment targets and to provide personalized care for patients with cerebrovascular disease. Intracerebral hemorrhage (ICH) is one of the most devastating cerebrovascular diseases with no effective treatment. To date, imaging markers of ICH risk- stratification and outcome prediction have been subjective and descriptive in nature, leaving a large gap for automated assessment of imaging feautres embedded in medical images. Preliminary results by Dr. Payabvash have demonstrated the feasibility of a research plan to apply automated feature extraction pipelines and machine learning algorithms to harness the information in medical images for early risk-stratification and identification of potential treatment targets in ICH. In this proposal, Dr. Payabvash will use detailed clinical and imaging data of 3,991 patients from NIH-funded clinical trials, online archives, and institutional registries at Yale, Tufts, and University College of London. He will apply machine-learning algorithms to identify those imaging features of brain hemorrhage on baseline head CT scan that are related to symptom severity at presentation (aim 1). Then, he will use imaging features of hemorrhage to identify those patients who are at risk for early expansion of hematoma (aim 2a), or surrounding edema (aim 2b). These two “modifiable” indicators of poor outcome are considered potential treatment targets in ICH patients. Finally, he will combine admission clinical information and imaging features to build a risk-stratification tool for long-term outcome prediction (aim 3). Under the expert mentorship of Dr. Kevin Sheth (Chief of Neurocritical Care), Dr. Todd Constable (Director of MRI Research), and Dr. Ronald Coifman (Professor of Mathematics), this K23 award will allow Dr. Payabvash to (1) identify and address the most pressing issues in cerebrovascular disease with innovative neurogaming tools; (2) gain expertise in advanced statistical analysis of brain scans; and (3) expand his knowledge in machine learning and computer vision for assessment of medical images. Dr. Payabvash will receive didactic training in neuroimaging statistical analysis, machine learning, deep neural networks, and computer vision. The proposed research and career development plans draw on the wealth of resources available at Yale, including a Regional Coordinating Center for the NIH StrokeNet, the Center for Research Computing; High Performance Computing services, and cutting-edge image processing and analysis infrastructure. At the conclusion of this award period, Dr. Payabvash will be well-positioned to become an independently-funded investigator conducting high-quality research in advanced neuroimaging techniques and analysis aimed at improving the care of patients with cerebrovascular disease.