Glucdict - Wearable Sensors and CGM

The Glucdict dataset consists on three types of data which arrive from three different sources:\ (1) blood glucose measurements from a Continuous Glucose Monitoring device.\ (2) a smartwatch device and its sensors’ measurements, which include heart rate, step detector and some motion sensors.\ (3) user activities, such as eating or drinking, which the user documents manually using a dedicated app installed on their smartphone.The dataset included 12 participants along 10 days. They were asked to wear the ”Mobvoi TicWatch Pro 2020” smartwatch on their right hand, the DEXCOM G6 CGM device, and their android smartphone.\ In addition, the activity types and duration were manually recorded by the participants each time they ate or drank using the ”GlucoDataSaver” app. The described experiment was approved by Human Subjects Research Committee of Ben-Gurion University.The recruitment period for this study spanned from October 26, 165 2021, to February 7, 2022. Each participant signed a consent document for the experiment which described the course of the experiment and its purpose.The participants were mostly students – 9 of 12, with age range between 22 and 63. The weight range was between 53 and 87 kg. The height range in centimeters was between 162 and 185. The gender distribution was 5 females and 7 males. One of the participants was a T1D.The dataset structured in the following way:Sensor IDs mapping:ACCELEROMETER = 1STEP_COUNTER = 19GYROSCOPE = 4MAGNETIC_FIELD = 2GRAVITY = 9LINEAR_ACCELERATION = 10ROTATION_VECTOR = 11EARTH_ACC = 99HEART_RATE = 21STEP_DETECTOR = 18There is a folder for each user (User1, ..., User12):Each user has 3 folders:Glucose: contains a csv file with 5 columns - index, timestamp, glucose value(mg/dL), transmitter and transmitter ID. Each row represents a glucose measurement (every 5 minutes).Phone: contains two folders - Activities and Sensors.Activities: contains csv files with manually recorded activities - mostly eat and drink and their timestamps.Sensors: contains csv files with the sensors data - each row consists a sensor ID, timestamp and values.Watch: contains csv files with the sensors data - same format as the phone's sensors.Glucdict数据集包含三类源自不同采集源的数据：(1) 连续血糖监测（Continuous Glucose Monitoring, CGM）设备采集的血糖测量值；(2) 智能手表设备及其传感器的测量数据，涵盖心率、步数检测器及部分运动传感器数据；(3) 用户手动通过智能手机上的专用应用程序记录的饮食饮水等用户活动。 该数据集共招募12名受试者，采集周期为10天。受试者被要求在右手佩戴"Mobvoi TicWatch Pro 2020"智能手表、DEXCOM G6 CGM设备，并随身携带安卓智能手机。此外，受试者每次饮食饮水时，均需通过"GlucoDataSaver"应用手动记录活动类型与持续时长。本实验已通过本-古里安大学人类受试者研究委员会的伦理审批。 本研究的招募周期为2021年10月26日至2022年2月7日。每名受试者均签署了实验知情同意书，内容涵盖实验流程与研究目的。 受试者群体以学生为主：12人中共有9名学生，年龄分布为22岁至63岁；体重范围为53kg至87kg；身高范围为162cm至185cm；性别分布为5名女性、7名男性。其中1名受试者为1型糖尿病（Type 1 Diabetes, T1D）患者。 该数据集的组织结构如下： 传感器ID映射表： 加速度计（ACCELEROMETER）= 1 计步器（STEP_COUNTER）= 19 陀螺仪（GYROSCOPE）= 4 磁场传感器（MAGNETIC_FIELD）= 2 重力传感器（GRAVITY）= 9 线性加速度传感器（LINEAR_ACCELERATION）= 10 旋转矢量传感器（ROTATION_VECTOR）= 11 地球加速度传感器（EARTH_ACC）= 99 心率传感器（HEART_RATE）= 21 步数检测器（STEP_DETECTOR）= 18 每名用户对应一个独立文件夹（User1至User12），每个用户文件夹下包含3个子文件夹： 1. Glucose：内含1个CSV文件，包含5列数据——索引、时间戳、血糖值（单位：mg/dL）、发射机信息及发射机ID。每一行代表一次血糖测量值，采集间隔为每5分钟1次。 2. Phone：内含两个子文件夹——Activities（活动记录）与Sensors（传感器数据）。 - Activities：内含CSV文件，存储手动记录的活动数据，主要为饮食饮水活动及其对应时间戳。 - Sensors：内含CSV文件，存储传感器采集数据，每行数据包含传感器ID、时间戳及传感器数值。 3. Watch：内含CSV文件，存储传感器采集数据，格式与手机端传感器数据完全一致。