SCH: Artificial Intelligence enabled multi-modal sensor platform for at-home health monitoring of patients

急性肾损伤(AKI)是一种常见的医学问题，与健康状况不佳有关。 幸存者的结果，包括死亡率增加和再次入院。尽管他们的风险很高 在AKI发作后，只有一小部分(10%)的患者接受专门的肾病医生的随访。 随访率低的原因是缺乏明确的指导方针，以及部分患者不情愿 由于医院疲劳、出行时间长以及不愿在医院增加更多医生等原因 护理团队。为了解决AKI幸存者护理方面的缺口，我们提出了一种启用人工智能(AI)的 居家使用的多模式传感器(MUSE)平台，可以自动监测患者的健康，执行 评估AKI复发的风险，并提醒患者寻求专门护理。MUSE包括-1) 用于测量尿液中生物标志物(肌酐、尿素、酸碱度和乳酸)浓度的比色尺； 2)近场通信(NFC)供电的可伸展、无电池的单导联心电图仪 心血管并发症后记录心电图的皮肤贴片是AKI复发的有力预测因素；3) 支持人工智能的移动应用程序，它获取传感器数据(从尿样和心电)，并运行设备上的深度学习融合AI模型，以将传感器数据和患者医疗记录(过去的共病)结合起来 和人口统计数据)，以实现精确和个性化的预测。我们将利用智能手机的功能 用于几个关键任务-a)使用内置摄像头捕获仪表板传感器的图像；b)充当NFC读取器 用于心电补丁；c)无需网络即可在板载运行计算机视觉和人工智能算法 连接，并在本地加密患者数据。人工智能模型将进行大规模的培训和验证 从梅奥诊所医院的患者那里收集的追溯数据集，传感器系统功能将 在Mayo对20名成年参与者(10名健康患者和10名AKI患者)进行的观察性研究证实了这一点。 这项拟议的研究有可能推动下一代智能汽车的创新 可穿戴设备，融合多模式传感器数据和电子病历，早期检测潜在健康状况 问题的精确度很高。提案目标的成功实现将为未来使用我们的传感器平台进行大规模临床试验铺平道路。

Acute kidney injury (AKI) is a commonly encountered medical problem that is associated with poor health outcomes in survivors, including increased mortality and re-admission to the hospital. Despite their high-risk status, only a small fraction (<10%) of patients receive specialized nephrologist follow-up after AKI episode. The low rate of follow-up care is due to lack of clear guidelines as well as reluctance on part of patients due to several reasons such as hospital fatigue, long travel times and unwillingness to add more doctors to the care team. To address the gap in care for AKI survivors, we propose an artificial intelligence (AI) enabled, MUlti-modal SEnsor (MUSE) platform for at-home use that can monitor patient health automatically, perform risk assessment for AKI recurrence, and alert the patient to seek specialized care. MUSE comprises of – 1) a colorimetric dipstick for capturing concentration of bio-markers (creatinine, urea, pH and lactate) in urine; 2) a near-field communication (NFC) powered stretchable, battery-less, single-lead electrocardiogram (ECG) skin patch that records ECG since cardiovascular complications is a strong predictor for AKI recurrence; 3) an AI-enabled mobile application that acquires sensor data (from urine sample and ECG) and runs an on-device deep learning fusion AI model to combine sensor data and patient medical record (past co-morbidities and demographics) for precision and personalized predictions. We will harness capabilities of smartphone for several key tasks - a) capture images of the dipstick sensor with built-in camera; b) act as NFC reader for ECG patch; c) run the computer vision and AI algorithms natively on-board without requiring network connection, and encrypt patient data locally. The AI model will be trained and validated on a large retrospective dataset collected from patients at Mayo Clinic Hospital, and the sensor system functionality will be validated with an observational study on 20 adult participants (10 healthy and 10 AKI patients) at Mayo. The proposed research has the potential to drive innovations in producing the next generation of intelligent wearables that performs fusion of multi-modal sensor data and EMR for early detection of underlying health issues with high accuracy. A successful realization of the proposal aims will pave the way for a future, large-scale clinical trial with our sensor platform.