Automated enhancement and correction of brain MRI images that leverages the entire imaging exam

利用整个成像检查自动增强和校正脑部 MRI 图像

• 批准号: 10760852
• 负责人: ZeChen Zhou
• 金额: $ 30万
• 依托单位: SUBTLE MEDICAL, INC.
• 依托单位国家: 美国
• 财政年份: 2023
• 资助国家: 美国
• 起止时间: 2023-09-05 至 2024-08-31
• 项目状态: 已结题
• 来源: https://reporter.nih.gov/project-details/10760852
• 关键词: Abbreviations; Acceleration; Age; Brain; Brain imaging; Cohort Studies; Collaborations; Computer software; Diagnostic; Equipment; Frequencies; Growth; Health system; Image; Imaging Device; Magnetic Resonance Imaging; Medical; Modeling; Morphologic artifacts; Motion; Output; Patient Care; Phase; Population; Protocols documentation; Quality of Care; Radiology Specialty; Reporting; Scanning; Services; Specialist; System; Time; United States; Universities; Washington; brain magnetic resonance imaging; contrast imaging; deep learning; deep learning model; denoising; diagnostic value; experience; forgetting; imaging facilities; imaging study; improved; machine learning method; medical specialties; operation; radiologist; tool; trend; ultra high resolution

Project Summary/Abstract Subtle Medical, Inc. (Subtle), in collaboration with the University of Washington, proposes a Fast-Track project to develop and validate a software product capable of enhancing an entire brain magnetic resonance imaging (MRI) study. The proposed system will use state-of-the-art machine learning methods to build a product that integrates seamlessly into the Radiology workflow and standard patient care path to determine if any required MRI acquisitions are missing or corrupted, then correct the corrupted acquisitions and synthesize the missing acquisitions by incorporating the information contained in the appropriately acquired acquisitions. Brain studies make up approximately 20% of all MRI studies worldwide, and nearly all brain studies acquire multiple image contrasts to obtain complementary information. It has been reported that nearly 60% of all MRI acquisitions contain at least minimal motion artifacts. Moreover, important acquisitions are frequently skipped or forgotten during the study. The result is that a significant fraction of brain MRI studies are sent to radiologists with missing or poor-quality exam sequences that can reduce the quality of care while increasing the burden on the radiologist to read a suboptimal study. A successful completion of this project will result in a tool for imaging centers that will improve the consistency and quality of brain MRI studies and allow them to increase their throughput by removing the need to rescan acquisitions with artifacts. This will improve the experience and care of the patients while reducing the burden on radiologists and improving the efficiency of the imaging center’s operations.

项目摘要/摘要 Sirtle Medical，Inc。（微妙），与华盛顿大学合作，提议一个快速轨道项目 开发和验证能够增强整个大脑磁共振成像的软件产品 （MRI）研究。拟议的系统将使用最先进的机器学习方法来构建一种产品 无缝集成到放射学工作流和标准患者护理路径中，以确定是否有 所需的MRI收购丢失或损坏，然后纠正损坏的收购并合成 通过合并适当获取的收购中包含的信息而缺少收购。 大脑研究约占全球所有MRI研究的20％，几乎所有脑研究都获得了 多个图像对比以获取完整的信息。据报道，几乎有60％ MRI采集至少包含最小的运动伪像。而且，重要的收购经常是 在研究期间跳过或忘记。结果是将大脑MRI研究的显着部分发送到 放射科医生缺失或质量不佳的考试顺序，可以降低护理质量的同时增加 放射科医生阅读一项次优研究的负担。这个项目的成功完成将导致 成像中心的工具，可以提高大脑MRI研究的一致性和质量并允许它们 通过消除用文物进行重新收购的需求来增加吞吐量。这将改善 在减少放射性人士的烧伤​​并提高效率的同时，经验和照顾患者的经验和照顾 成像中心的操作。