An Integrative Radiogenomic Approach to Design Genetically-Informed Image Biomarker for Characterizing COPD

设计用于表征 COPD 的遗传信息图像生物标志物的综合放射基因组方法

• 批准号: 10866646
• 负责人: Kayhan Batmanghelich
• 金额: $ 47.82万
• 依托单位: BOSTON UNIVERSITY (CHARLES RIVER CAMPUS)
• 依托单位国家: 美国
• 财政年份: 2018
• 资助国家: 美国
• 起止时间: 2018-05-01 至 2025-04-30
• 项目状态: 未结题
• 来源: https://reporter.nih.gov/project-details/10866646
• 关键词: Integrative; Radiogenomic; Approach; Design; Genetically

! Abstract Chronic Obstructive Pulmonary Disease (COPD) is one of the leading causes of death worldwide with a devastating socio-economic burden impacting more than three million individuals per year in the US. The primary environmental risk factor in the susceptible population is smoking, which causes an exaggerated inflammatory response. However, many factors including several genetic risk variants substantially influence the susceptibility. Twin-based studies show that families with emphysema have a higher risk for the disease. The two different major phenotypes of COPD are small airway remodeling (airway disease) and alveolar destruction (emphysema). Although these two major phenotypes result in a similar deficiency in global lung function, the relationship between them is complicated and likely involves feedback mechanisms. Developing an objective method to characterize lung phenotypes is critical since treatment candidates vary based on phenotype. Measurements from High-Resolution Computed Tomography (HRCT) images are increasingly used to describe COPD since they can quantitatively describe the contribution of the phenotypes. To discover the genetic risk variants, Genome-Association Studies (GWAS) have focused on either the physiological lung function or a simple threshold-based measurement from lung CT, neither of which fully characterizes phenotypic subtypes or the distribution pattern of disease. The proposed studies will take advantage of the rich image and genetic data jointly to build a genetically-informed imaging biomarker to characterize each patient. For each patient, our method summarizes the CT image to a vector representation that accurately describes the severity of the disease. Also, a method to link the representation back to the genetic risk variants will be developed. If successful, these methods can be used to monitor the efficacy of treatment or progression of the disease using imaging data. Successful execution of the second aim will result in better understanding of the etiology of different disease subtypes and discovery of novel genetic pathways that could be used as potential drug targets. Furthermore, the patient representation enables the use of image data to construct a more powerful model to predict the so-called acute exacerbation event. Predicting the exacerbations is clinically important since they cause further damage to the lung. In Aim 1, we develop and implement a novel image biomarker that is mutually informed by imaging and genetic data from each patient. Our statistical method in Aim 2 elucidates the underlying genetic pathways behind the abnormal anatomical variations explained by the biomarker. We validate our method on data from 10,300 patients in the COPDGene dataset. !

! 摘要 慢性阻塞性肺疾病（COPD）是全球主要的死亡原因之一， 毁灭性的社会经济负担，每年影响美国300多万人。的 易感人群中的主要环境风险因素是吸烟，这会导致过度的 炎症反应。然而，包括几种遗传风险变异在内的许多因素会显著影响 易感性。基于双胞胎的研究表明，患有肺气肿的家庭患肺气肿的风险更高。 COPD的两种不同的主要表型是小气道重塑（气道疾病）和肺泡灌洗液（肺泡灌洗液）。 肺气肿（emphysema）。虽然这两种主要的表型导致了全球肺组织中类似的缺陷， 功能，它们之间的关系是复杂的，可能涉及反馈机制。发展中 表征肺表型的客观方法是至关重要的，因为治疗候选者基于以下因素而变化： 表型来自高分辨率计算机断层扫描（HRCT）图像的测量越来越多地被用于测量图像的分辨率。 用于描述COPD，因为它们可以定量描述表型的贡献。发现 基因组关联研究（GWAS）的重点是生理性肺 功能或来自肺部CT的简单基于阈值的测量，两者都不能完全表征 表型亚型或疾病的分布模式。拟议中的研究将利用富人 以建立遗传学上知情成像生物标记来表征每个患者。 对于每个患者，我们的方法将CT图像总结为准确描述 疾病的严重性。此外，将表示与遗传风险变体联系起来的方法将是 开发如果成功，这些方法可用于监测治疗的疗效或进展。 利用影像学数据诊断疾病。成功地实现第二个目标将使人们更好地了解 不同疾病亚型的病因学和新的遗传途径的发现，可用于潜在的 药物靶点此外，患者表示使得能够使用图像数据来构建更好的诊断。 预测所谓急性加重事件的强大模型。预测急性加重在临床上 这很重要，因为它们会对肺部造成进一步的损伤。 在目标1中，我们开发并实现了一种新的图像生物标志物，该生物标志物通过成像和 每个病人的基因数据。我们在目标2中的统计方法阐明了潜在的遗传途径 生物标记物所解释的异常解剖学变异的背后。我们验证了我们的方法的数据， COPDGene数据集中的10,300名患者。 !

项目成果

Can contrastive learning avoid shortcut solutions?

• 发表时间: 2021-06
• 期刊: Advances in neural information processing systems
• 作者: Joshua Robinson;Li Sun;Ke Yu;K. Batmanghelich;S. Jegelka;S. Sra

Generative-Discriminative Complementary Learning

• DOI: 10.1609/aaai.v34i04.6126
• 发表时间: 2019-04
• 期刊: Proceedings of the ... AAAI Conference on Artificial Intelligence. AAAI Conference on Artificial Intelligence
• 作者: Yanwu Xu;Mingming Gong;Junxiang Chen;Tongliang Liu;Kun Zhang;K. Batmanghelich

Augmentation by Counterfactual Explanation - Fixing an Overconfident Classifier.

• DOI: 10.1109/wacv56688.2023.00470
• 发表时间: 2023-01
• 期刊: IEEE Winter Conference on Applications of Computer Vision. IEEE Winter Conference on Applications of Computer Vision
• 作者: Singla, Sumedha;Murali, Nihal;Arabshahi, Forough;Triantafyllou, Sofia;Batmanghelich, Kayhan
• 通讯作者: Batmanghelich, Kayhan

Improving clinical disease subtyping and future events prediction through a chest CT-based deep learning approach.

• DOI: 10.1002/mp.14673
• 发表时间: 2021-03
• 期刊: Medical physics
• 影响因子: 3.8
• 作者: Singla S;Gong M;Riley C;Sciurba F;Batmanghelich K
• 通讯作者: Batmanghelich K

Deep Learning Prediction of Voxel-Level Liver Stiffness in Patients with Nonalcoholic Fatty Liver Disease.

• DOI: 10.1148/ryai.2021200274
• 发表时间: 2021-09
• 期刊: Radiology. Artificial intelligence
• 作者: Brian L Pollack;K. Batmanghelich;Stephen S Cai;E. Gordon;Stephen Wallace;R. Catania;Carlos Morillo-Hernandez;A. Furlan;A. Borhani