Smart Toilet with Artificial Intelligence for Remote Management of Inflammatory Bowel Disease

摘要 炎症性肠病（IBD）影响三百万美国成年人，并且由于慢性，需要 终身医疗，对患者的生活质量产生负面影响，并导致大规模医疗保健 成本病程的特点是缓解和突然发作，因此疾病 通过药物和饮食干预进行管理需要经常监测。侵入性较小 与内窥镜检查相比，粪便生物标志物钙卫蛋白的纵向跟踪越来越多地用于 达到并维持缓解。通过基于应用程序的日记进行远程疾病管理， 肠运动来评估疾病活动性和在家测量钙卫蛋白， 发现一致实施时有效;然而，患者依从性低至30%。 我们提出了一种技术，使自动和无缝监测排便 和钙卫蛋白，从而大大增加患者对跟踪方案的依从性。的 创新是一个智能取样厕所系统：一个改造成传统厕所的工程入口， 一种马桶，可在冲洗后仔细检查粪便外观并实现免持粪便 取样分析。该系统已在实验室中用人类粪便进行了演示 标本的粪便形式的范围，包括松散的粪便。人工智能（AI） 提出了一种粪便图像分析和分类的方法。 本研究的目的是证明通过以下方法现场测量钙卫蛋白的可行性： 智能取样马桶，一个供人类使用的原型系统， 形态分析特定目标1是一项使用商业侧流测定的实验室研究， 评估通过粪便采样的钙卫蛋白测量的灵敏度和特异性 设计的入口在加标研究后，我们将测量临床标本中的钙卫蛋白， 通过与杜克大学胃肠病学IBD诊所合作， 在具体目标2中，我们将在实验室外部署一个智能马桶原型， 人类受试者，以完善粪便形态分类的AI算法。我们将评估 来自使用厕所的健康受试者的样本和来自冲洗的临床样本 从IBD患者中获得。 本研究中获得的数据将成为智能马桶系统第二阶段研究的基础 由杜克IBD诊所安装供患者使用。最终，我们设想智能马桶 系统安装在家里，有利于IBD疾病管理和提高病人的质量， 通过消除粪便数据访问的障碍并改善临床结果，

ABSTRACT Inflammatory Bowel Disease (IBD) affects three million US adults and, due to chronicity, requires lifelong medical care, negatively impacts patient quality of life, and results in large healthcare costs. The disease course is characterized by remissions and flare-ups, thus disease management by medication and dietary interventions requires frequent monitoring. Less invasive than endoscopy, longitudinal tracking of the fecal biomarker calprotectin is increasingly used to achieve and maintain remission. Remote disease management through app-based diaries of bowel movements to assess disease activity and at-home measurement of calprotectin have been found effective when implemented consistently; however, patient adherence is as low as 30%. We propose a technology that enables automatic and seamless monitoring of bowel movements and calprotectin, thereby greatly increasing patient adherence to tracking regimens. The innovation is a Smart Sampling Toilet system: an engineered portal retrofitted to a conventional toilet that, discreetly and post-flush, inspects stool appearance and enables hands-free stool sampling for analysis. The system has been demonstrated in the laboratory with human stool specimens across the range of stool forms, including loose stools. An artificial intelligence (AI) method has been developed for stool image analysis and classification. The objective of this study is to demonstrate feasibility of onsite measurement of calprotectin by the Smart Sampling Toilet, a prototype system for use by human subjects that integrates stool morphology analysis. Specific Aim 1 is a benchtop study using commercial lateral flow assays to assess sensitivity and specificity of calprotectin measurements from stool sampled through the engineered portal. After a spiking study, we will measure calprotectin on clinical specimens from patients with IBD through a collaboration with the Duke University Gastroenterology IBD Clinic. In Specific Aim 2, we will deploy a Smart Toilet prototype outside of the laboratory for use by human subjects to refine the AI algorithm for stool morphology classification. We will evaluate specimens both from healthy subjects using the toilet and from flushing clinical specimens obtained from IBD patients. The data obtained in this study will be the basis for a Phase 2 study with the Smart Toilet system installed for use by patients through the Duke IBD clinic. Ultimately, we envision Smart Toilet systems installed in the home to benefit IBD disease management and improve patient quality of life by removing barriers to stool-data access and enabling improved clinical outcomes.