SBIR Phase I: A Real-time Artificial Intelligence Dysplasia Detection (AIDD) system for Barrett's Esophagus and Early Esophageal Cancer

SBIR 第一阶段：针对巴雷特食管和早期食管癌的实时人工智能发育不良检测（AIDD）系统

• 批准号: 1843975
• 负责人: Andrew Ninh
• 金额: $ 22.5万
• 依托单位: DOCBOT, INC.
• 依托单位国家: 美国
• 项目类别: Standard Grant
• 财政年份: 2019
• 资助国家: 美国
• 起止时间: 2019-02-01 至 2019-09-30
• 项目状态: 已结题
• 来源: https://www.nsf.gov/awardsearch/showAward?AWD_ID=1843975&HistoricalAwards=false
• 关键词: SBIR; Phase; Real; time; Artificial

The broader impact/commercial potential of this Small Business Innovation Research (SBIR) Phase I project is to, through development of a Real-time Artificial Intelligence algorithm, bring expertise in endoscopic imaging interpretation to physicians taking care of patients at risk for esophageal cancer, to detect precancer earlier, and lower the cost of healthcare. Esophageal cancer is one of the fastest growing cancers in the US with a 6 fold increase in incidence in the last 4 decades. Most endoscopists are not trained well to detect pre-cancer in the esophagus as 5% have received dedicated advanced imaging training to identify dysplasia (advanced precancer) within Barrett's Esophagus. There are well-proven minimally invasive endoscopic treatments for dysplasia and early esophageal cancer that are 95% effective in providing cure to the patient. On the other hand, when cancer is caught at a later stage the only option is removal of the esophagus with surgery in combination with expensive chemotherapy and radiation. Creating this real-time Artificial Intelligence system will assist endoscopists to detect precancer earlier, prevent this deadly cancer and expand access to esophageal cancer screening to the community at large and also to the underserved community, who typically have fewer well-trained physicians.This Small Business Innovation Research (SBIR) Phase I project will develop the first system that can detect dysplasia in Barrett's Esophagus in real-time during endoscopy. This proposed research will expand on the company's preliminary work in detecting dysplastic lesions in colonoscopy and applying transfer learning methods to develop the first ever dysplasia-detection algorithm for upper endoscopy. It will initially be trained using static, annotated images, and will expand on the company's software platform to be able to eventually process video feeds, in real-time, from device makers without frame-drop or lag. The company has begun engaging with endoscopy OEMs to define metrics for success in running real-time algorithms.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.

这个小型企业创新研究（SBIR）第一阶段项目的更广泛的影响/商业潜力是，通过开发实时人工智能算法，将内窥镜成像解释的专业知识带给医生，以照顾有食管癌风险的患者，更早地检测癌前病变，并降低医疗保健成本。 食管癌是美国增长最快的癌症之一，在过去40年中发病率增加了6倍。大多数内窥镜医生没有接受过良好的培训来检测食管中的癌前病变，因为5%的内窥镜医生接受过专门的高级成像培训来识别巴雷特食管中的异型增生（晚期癌前病变）。对于异型增生和早期食管癌，有充分证明的微创内镜治疗方法，在为患者提供治愈方面有95%的有效性。另一方面，当癌症在后期被发现时，唯一的选择是手术切除食管，并结合昂贵的化疗和放疗。创建这种实时人工智能系统将有助于内窥镜医生更早地发现癌前病变，预防这种致命的癌症，并将食管癌筛查的范围扩大到整个社区以及服务不足的社区，这个小企业创新研究（SBIR）第一阶段项目将开发第一个可以在真实的中检测巴雷特食管发育不良的系统-内窥镜检查期间的时间。这项拟议的研究将扩大该公司在结肠镜检查中检测发育异常病变的初步工作，并应用迁移学习方法开发首个用于上消化道内窥镜检查的发育异常检测算法。它最初将使用静态的注释图像进行训练，并将在公司的软件平台上扩展，最终能够实时处理来自设备制造商的视频，而不会出现帧丢失或延迟。该公司已开始与内窥镜OEM合作，以定义运行实时算法的成功指标。该奖项反映了NSF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。