Use of Deep Learning Algorithms to Enable Evaluation of the Determinants and Outcomes of Hepatic Steatosis, by HIV Status

使用深度学习算法根据 HIV 状态评估肝脂肪变性的决定因素和结果

• 批准号: 10677782
• 负责人: Jessie Torgersen
• 金额: $ 16.79万
• 依托单位: UNIVERSITY OF PENNSYLVANIA
• 依托单位国家: 美国
• 财政年份: 2022
• 资助国家: 美国
• 起止时间: 2022-08-09 至 2027-04-30
• 项目状态: 未结题
• 来源: https://reporter.nih.gov/project-details/10677782
• 关键词: Address; Aging; Alcohol abuse; Alcohol consumption; Algorithms; Artificial Intelligence; Bioinformatics; Biometry; CD4 Lymphocyte Count; Cardiovascular system; Caring; Cells; Cirrhosis; Classification; Clinical; Cohort Studies; Computed Tomography Scanners; Computer Assisted; Computerized Medical Record; Data; Development; Diabetes Mellitus; Diagnosis; Dyslipidemias; Electronic Health Record; Epidemiology; Evaluation; Fatty Liver; Future; General Population; HIV; HIV Infections; Hepatic; Hepatocyte; Hepatology; Hypertension; Image; Informatics; Intervention; Knowledge; Liver; Liver diseases; Magnetic Resonance Imaging; Manuals; Measurement; Mediating; Medicine; Mentors; Mentorship; Metabolic dysfunction; Methodology; Methods; Morbidity - disease rate; Obesity; Outcome; Outcome Study; Pathway interactions; Patients; Persons; Population; Primary carcinoma of the liver cells; Prospective, cohort study; Research; Research Personnel; Resources; Risk; Risk Factors; Risk Reduction; Sample Size; Sampling; Scanning; Science; Selection Bias; Selection for Treatments; Severities; Surrogate Markers; Time; Training; Triglycerides; United States; Veterans; Veterans Health Administration; Viral; Viral hepatitis; Viremia; Work; X-Ray Computed Tomography; abdominal CT; antiretroviral therapy; artificial intelligence method; clinical care; co-infection; cohort; deep learning; deep learning algorithm; elastography; follow-up; interest; large datasets; liver biopsy; liver inflammation; liver transplantation; mortality; multidisciplinary; novel; obesogenic; prevent; preventive intervention; radiological imaging; radiologist; repository; screening; tool

Hepatic steatosis, defined by >5% hepatocyte triglyceride content, may be potentiated in people with HIV (PWH) through viral-mediated mechanisms or metabolic dysfunction associated with antiretroviral therapy (ART). However, the epidemiology of hepatic steatosis remains unclear among PWH, primarily because studies have been limited to small patient samples that ascertained steatosis via specialized radiographic methods or liver biopsy. Since liver disease is a leading cause of morbidity and mortality among PWH, it is critically important to identify the determinants and consequences of hepatic steatosis in this group. Such studies will inform interventions and management strategies to mitigate HIV-specific steatosis mechanisms and its consequences, particularly hepatic decompensation and hepatocellular carcinoma (HCC). Recent advances in artificial intelligence have facilitated the development of automated computer-aided liver assessment to determine the presence and severity of hepatic steatosis within noncontrast abdominal computed tomography (CT) scans. The 8utomatic ,!:iver 8ttenuation Region-Of-Interest-based Measurement (ALARM) is a deep learning tool previously developed for the identification of moderate-to-severe hepatic steatosis. Preliminary studies conducted by the applicant demonstrate the high accuracy of ALARM compared to manual radiologist review across multiple centers and CT scanners, including within the Veterans Health Administration. To address the knowledge gaps of existing studies, this proposal will first establish a cohort of over 40,000 PWH and people without HIV (PWOH) in the Veterans Aging Cohort Study (VACS) who underwent noncontrast abdominal CT imaging for any indication in the context of clinical care between 2002- 2020. The VACS, an ongoing national prospective cohort study of PWH and PWOH across the United States, includes access to electronic health record data, including image files of CT scans. The ALARM tool will be applied to this repository of radiographic images to objectively classify the presence or absence of moderateto- severe hepatic steatosis. The research plan aims to: 1) identify the HIV-specific determinants associated with hepatic steatosis among PWH, 2) define how traditional determinants of steatosis differ by HIV status, and 3) determine the risk of liver complications associated with steatosis in PWH and how this risk differs by HIV status. The findings from these studies will inform interventions to prevent and mitigate the development of hepatic steatosis among persons with HIV, which will help lower the risk of liver complications and prolong survival in this population. This project will bring together a mentoring team of nationally recognized researchers and provide time for coursework and training in advanced epidemiology, biostatistics, informatics, artificial intelligence, hepatology, and HIV medicine that are needed to establish the applicant as an independent investigator in the field of HIV-related liver diseases.

肝脂肪变性为5％的肝细胞甘油三酸酯含量，艾滋病毒患者可能会增强 （PWH）通过病毒介导的机制或与抗逆转录病毒疗法相关的代谢功能障碍 （艺术）。但是，肝脂肪变性的流行病学在PWH中尚不清楚，主要是因为 研究仅限于小型患者样品，这些样本通过专门的射线照相确定了脂肪变性 方法或肝活检。由于肝病是PWH中发病率和死亡率的主要原因，因此 对于确定该组肝脂肪变性的决定因素和后果至关重要。这样的 研究将为干预措施和管理策略提供信息，以减轻HIV特异性脂肪变性机制和 它的后果，尤其是肝功能补偿和肝细胞癌（HCC）。 人工智能的最新进展有助于自动计算机辅助肝脏的发展 评估以确定非对比度腹部肝脂肪变性的存在和严重程度 计算机断层扫描（CT）扫描。 8局部！ （警报）是一种以前开发的深度学习工具，用于识别中度至重度肝 脂肪变性。申请人进行的初步研究证明了警报的高度准确性 在多个中心和CT扫描仪中，包括退伍军人健康中的手动放射科医生审查 行政。为了解决现有研究的知识差距，该建议将首先建立一组 超过40,000个PWH和没有艾滋病毒的人（PWOH）在老年人群研究（VACS）中 在2002年之间进行临床护理的任何适应症进行了非对比度的腹部CT成像 2020。 包括访问电子健康记录数据，包括CT扫描的图像文件。警报工具将是 应用于该放射线图像的存储库，以客观地将其存在或不存在 严重的肝脂肪变性。研究计划的目的是：1）确定与HIV特异性决定因素相关的决定因素 PWH中的肝脂肪变性，2）定义脂肪变性的传统决定因素如何因艾滋病毒状况而不同，而 3）确定与PWH中与脂肪变性相关的肝脏并发症的风险以及这种风险因HIV的不同而差异 地位。这些研究的发现将为您提供干预措施，以防止和减轻 艾滋病毒患者之间的肝脂肪变性，这将有助于降低肝脏并发症的风险并延长 在这个人群中的生存。该项目将汇集一个由全国认可的指导团队 研究人员并为高级流行病学，生物统计学，信息学，信息提供时间进行课程和培训时间 人工智能，肝病学和艾滋病毒医学是建立申请人为 HIV相关肝病领域的独立研究者。