SBIR Phase I Topic 402: Artificial Intelligence-Aided Imaging for Cancer Prevention, Diagnosis, and Monitoring

SBIR 第一阶段主题 402：用于癌症预防、诊断和监测的人工智能辅助成像

• 批准号: 10347278
• 负责人: XUE FENG
• 金额: $ 5.5万
• 依托单位: CARINA MEDICAL, LLC
• 依托单位国家: 美国
• 财政年份: 2020
• 资助国家: 美国
• 起止时间: 2020-09-16 至 2021-06-15
• 项目状态: 已结题
• 来源: https://reporter.nih.gov/project-details/10347278
• 关键词: Address; Algorithms; Artificial Intelligence; Cervical lymph node group; Clinical; Complex; Computer software; Data; Detection; Diagnosis; Evaluation; Head and Neck Cancer; Human; Image; Interobserver Variability; Learning; Magnetic Resonance Imaging; Malignant Neoplasms; Measures; Modality; Monitor; Patient Care; Performance; Phase; Privatization; Reporting; Site; Small Business Innovation Research Grant; Software Framework; Visualization; automated segmentation; base; cancer diagnosis; cancer imaging; cancer prevention; clinical practice; clinically relevant; cloud based; cost; imaging modality; lymph nodes; novel; segmentation algorithm; usability

Image-based evaluation of lymph nodes is an essential step in cancer diagnosis, treatment and monitoring. Current clinical practice mostly uses qualitative or semi-quantitative measures in evaluation and thus suffers from inaccuracy due to intra- and inter-observer variability and increased human efforts. This becomes a more serious issue in head and neck cancers due to the large number of clinically relevant lymph nodes. In this project an AI-based automatic segmentation software will be developed for quantitative cervical lymph node evaluation to increase the accuracy and reduce the cost. However, there are a few challenges in developing and deploying such a software due to different clinical practices such as usage of different modalities (MRI and/or CT) and complex clinical workflow. To address these challenges, a novel AI algorithm that can handle the variability in imaging modalities and support incremental learning using site-specific data to enhance its robustness will be developed; a private-cloud-based software framework with high usability will then be developed to incorporate this algorithm and provide advanced visualization and reporting for clinical usage. This software will have high impact on all stages of patient care for head and neck cancers and can be further extended to other cancers.

基于图像的淋巴结评估是癌症诊断，治疗和监测的重要步骤。当前的临床实践主要在评估中使用定性或半定量措施，因此由于观察者内和观察者间的变异性和人类的努力增加而遭受不准确。由于大量临床相关的淋巴结，这成为头颈癌中更严重的问题。在此项目中，将开发基于AI的自动分割软件，以进行定量的宫颈淋巴结评估，以提高准确性并降低成本。但是，由于不同的临床实践，例如使用不同的方式（MRI和/或CT）以及复杂的临床工作流程，在开发和部署此类软件方面存在一些挑战。为了应对这些挑战，一种新型的AI算法可以使用特定于网站的数据来处理成像方式的变异性并支持增量学习以增强其鲁棒性；然后将开发具有高可用性的基于私有云的软件框架，以结合此算法并提供高级可视化和用于临床使用的报告。该软件将对头颈癌的患者护理的所有阶段产生很大的影响，并可以进一步扩展到其他癌症。