SCH:Artificial Intelligence for Contrast-Enhanced Imaging

肝脏肿瘤的病例及其相关费用在过去几年中一直在稳步增加。目前诊断肝脏肿瘤需要在做磁共振成像（MRI）时给患者注射化学造影剂。这些化学造影剂使用时间长，价格昂贵，而且对许多人来说有可怕的副作用。这个智能和互联健康（SCH）奖项汇集了一个多学科团队，包括来自计算机科学、生物医学工程和临床放射学的研究人员，共同开发一种人工智能生成的虚拟对比MRI，从而减少时间、成本和化学对比剂造成的发病率。该项目将为不同群体的学生和临床研究员提供机器学习、图像处理和医学成像方面的跨学科培训。此外，该项目将招收初高中学生（K-12外展），博士生和博士后。该项目提出了基于上下文对抗学习、不确定性和可靠性分析的新方法，以实现对对比度增强成像的基本理解和计算机建模。项目拟(1)研究“对比经济学”这一全新的语境特征范畴；(2)开发新的跨域上下文模型来检测、分类和量化病变，并合成与真实对比度增强成像具有同等诊断价值的虚拟对比度图像；(3)开发新颖的不确定性和可靠性分析，以获得最终用户的信任；(4)利用MRI对肝癌/肿瘤分类模型进行验证。该项目汇集了计算机科学、生物医学工程和临床放射学研究人员，并提出了一个综合的多学科教育和推广计划，以实现从这项工作中获得的知识的最广泛传播。该项目还提出了潜在的未来转移到实践（工业和诊所），并将进步广泛传播到图像处理、机器学习和医疗数据分析领域以外的广泛社区。该奖项反映了美国国家科学基金会的法定使命，并通过使用基金会的知识价值和更广泛的影响审查标准进行评估，被认为值得支持。

The instances of liver tumors and its associated cost has been increasing steadily in the last few years. Diagnosing liver tumors currently requires injecting patients with a chemical contrast agent while doing magnetic resonance imaging (MRI). These chemical contrast agents are time consuming to administer, expensive, and have morbid side effects for many individuals. This Smart and Connected Health (SCH) award brings together a multidisciplinary team, comprising researchers from computer science, biomedical engineering, and clinical radiology to develop an Artificial Intelligence generated virtual contrast MRI thereby reducing the time, cost, and morbidity due to the chemical contrast agent. This project will provide a diverse group of students and clinical fellows with interdisciplinary training in machine learning, image processing, and medical imaging. Additionally, this project will engage students from middle/high school students (K-12 outreach) to doctoral students and postdoctoral fellows.This project proposes novelties in contextual adversarial learning, uncertainty, and reliability-based analysis to enable fundamental understanding and computer modelling of contrast enhanced imaging. The project proposes to (1) investigate “contrasomics”, a brand new category of contextual features; (2) develop novel cross-domain contextual models to detect, classify, and quantify lesions, and to synthesize virtual contrast images that have equivalent diagnostic value with real contrast enhanced imaging; (3) develop novel uncertainty and reliability analysis to gain the trust of end users; (4) validate the models with liver cancer/tumour classification using MRI. This project brings together computer science, biomedical engineering, and clinical radiology researchers and proposes an integrated multi-disciplinary education and outreach program to achieve the broadest possible dissemination of the knowledge gained from this work. The project also proposes potential future transfer to practice (industry and clinic) and wide dissemination of advances to broad communities beyond the image processing, machine learning, and medical data analytics domains.This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.