A Rapid Carotid Atherosclerotic Disease Assessment System with 3D MRI

点击翻译按钮获取中文摘要

Abstract Stroke is a major cause of death and long-term disability worldwide. In the United States alone, there are roughly 0.8 million ischemic strokes each year, with an estimated cost of $34 billion. Among the various etiologies of stroke, carotid atherosclerotic disease (CAD) is a very common one. Accurate diagnosis of CAD will be important for effective treatment and prevention of stroke. Although carotid vessel wall (CVW) magnetic resonance imaging (MRI) has shown superior advantages in CAD assessment over other imaging techniques (e.g. ultrasound, CTA, etc.), its widespread in clinical use is slow due to several limitations, including lengthy and complex imaging procedure, highly demanding on technologists’ training and experiences and laborious lesion analysis, etc. Therefore, there is a clear need for a fast and automated image acquisition and analysis solution to improve the adoption in clinical settings. In this Phase I project, we plan to develop a rapid clinical CAD assessment solution, which includes a large coverage single contrast MR imaging procedure and a fully automated analysis software, named CASCADIA, with innovative image quality evaluation (IQE) and deep learning lesion analysis technologies. We also designed a new domain adaptation framework to ensure CASCADIA’s performance on images acquired from various imaging sites. This system is expected to complete clinical CAD assessment of a patient with unprecedented short time of 5 minutes which is 1/10 of the time in current clinical use. This solution will not only be an efficient assistant diagnosis tool for clinicians but also benefit patients with shorter imaging time and potential cost deduction. The output of Phase I product will include image quality evaluation, clinical measurements and atherosclerotic lesion classification. Development of more advanced analysis, such as lesion characterization and risk score, are planned in Phase II. We currently have access to more than 1000 3DMERGE scans acquired from previous research projects (a large portion of them with radiologists’ annotations), with which, in Phase I, we propose to: Aim 1: Develop the prototype software of CASCADIA, which includes: 1) Develop a streamlined CAD assessment workflow by combining IQE and lesion classification modules 2) Integrate clinical measurement functions, including stenosis, plaque burden and wall thickness Aim 2: Evaluate performance of the prototype software, which includes: 1) Evaluate the performance of IQE and lesion classification modules in CASCADIA 2) Install the proposed solution in a clinical imaging site and test with phantom This Phase I proposal is planned for six months with clearly defined timeline and milestones. The developed prototype will be deployed in clinical settings as a pre-market base for our long-term goal of full- scale commercial product development in a Phase II application.