greenteg-core-axillary-1Hz

Core body temperature measures from a single subject with a non-24 circadian rhythm disorder, aged 33 years old at the start of measurements, weight between 72 and 77.5 kg. <br> The core body temperature measurements were done non-invasively with the GreenTEG CORE Research device, attached to the side of a Polar Pro chest strap band, placed on the axillary/apically per the manufacturer's recommendations. The sensor was always placed on the right side of the chest. The device captures core body temperature and skin temperature every second (1Hz sampling rate). It can store up to 3.5 days of data and takes 3h30 to 7h for data download, hence the gaps, which are placed during circadian daytime whenever possible. This device is able to measure the core body temperature using dual heat flux method (DHFM) with an in-house AI algorithm that replaces the standard DHFM equation to obtain more reliable readings according to the manufacturer. At the time of this experiment, this was the first and only wearable thermometer using heat flux technology. <br> The measurements are much more reliable since January 2021 so prefer to analyze the data from this date onwards. The device was changed in January 2021 for a new one because the old one broke due to a mechanical pressure flaw in the design. Since January 2021, an experimental attachment design using velcro sticked vertically on the CORE and with a loop around the Polar Pro band was conceived by the author/subject to avoid this mechanical flaw, and this had the unexpected side effect of improving significantly the reliability and extent of the surface of the CORE being in contact with the skin, so that measures are much more reliable and systematically reflect the circadian rhythm. Usually, the circadian night can be detected simply as being periods with a temperature below 36.5°C with this new attachment system. <br> Along with the sleep diaries, this is the most reliable circadian rhythm monitoring marker so far from preliminary results. <br> Note on data columns: most columns are encrypted gibberish data. Only the core body temperature, skin temperature flux 1, and timestamp are accessible and should be processed, the other columns can and should be discarded. More precisely, only the following columns are available (from the manufacturer's documentation): <br> * time [UTC-OFS=+0100] Timestamp of the data point in the format ‘DD:MM:YYYY hh:mm:ss’ * timestamp [us] Timestamp in UNIX time in us (microseconds elapsed since January 1th 1970). * Hf_a0 Raw heat flux sensor signal of heat path A in ADC counts (unitless). * Temp_a0 [mC] Uncorrected temperature signal of heat path A in millidegree Celsius. * Cbt [mC] Core body temperature estimation output in millidegrees.<br> <br> Note: Cbt needs to be divided by 1000 to get the core body temperature value in celsius degrees. <br> The following columns are not available and can be dropped as they are filled with random values: hf_a1, Temp_a1, ax, ay, az, Battery_voltage [mV]. <br> Heat flux sensor A value can be obtained as follows: <br> Heat flux sensor A voltage [in uV] = hf_a0 [in counts] * 1.953125 uV<br> <br> Skin temperature for sensor A can be obtained as follows: <br> Skin Temperature A [in °C] = (temp_a0 [in mC] – T0off) /1000<br> <br> T0off is an offset that is defined once in the header of each CSV file. <br> There is a gap during January 2021, as this is the period the 1st device broke, before getting replaced by the manufacturer. The new dataset with the new device starts on 21th January 2021, with a new vertically oriented velcro attachment system devised by the author, which prevents the device from breaking and improves skin contact and hence data quality. It is recommended to analyze data after 21th January 2021 as it is likely the least noisy. <br> Each file represents one acquisition session of up to 3.5 days, hence they can be concatenated to study a longer time period. <br> Note: On 2021-05-29T21_25_00, an attempt to optimize further the disposition of the velcro was done by changing the orientation from vertical to diagonal, so that the CORE device was oriented on skin as a losange instead of a square, which would theoretically further increase skin contact and data quality, since the chest strap now pushes on the two corners in the horizontal middle, with much less chances that it gets stuck in a local equilibrium at either of the vertical corners. Unfortunately, the data quality was worse (the high and low phases were not correlated with empirical sleep-wake data anymore, there were no high nor low phases anymore), suggesting that the two heat flux sensors are vertically positioned in the middle line of the device (one temperature sensor at the top and one at the bottom), and hence that a vertical orientation is likely the most optimal orientation despite reduced skin contact compared to diagonal. Hence, on 2021-06-06T01_00_00, the disposition of the velcro band was reverted back to vertical. Only one session was acquired with the diagonal orientation, the file is named accordingly. <br> To postprocess the data for analysis, ensure that the first 30 min of each file / new data collection session is removed, as the sensor takes some time to converge to the true current body temperature. It will always start too high at first (because the equation for dual heat flux will show a high value when the sensor is cold when it was not worn), and then gradually lowers until it converges to the equilibrium which represents the current core body temperature. <br>

本数据集包含一名非24小时昼夜节律障碍（non-24 circadian rhythm disorder）受试者的核心体温测量数据，测量开始时受试者年龄为33岁，体重介于72至77.5千克之间。
核心体温测量采用非侵入式方式，使用GreenTEG CORE Research设备，该设备固定于Polar Pro胸带侧面，按照制造商建议放置于腋下/心尖部位，且传感器始终置于胸部右侧。该设备以每秒1次（1Hz采样率）的频率采集核心体温与皮肤温度数据，可存储最长3.5天的测量数据，数据下载耗时3.5小时至7小时，因此存在数据间隙，此类间隙尽可能安排在昼夜节律的日间时段。
该设备采用双热通量法（Dual Heat Flux Method, DHFM）进行核心体温测量，搭载自研人工智能算法替代标准DHFM方程，以获得更可靠的测量结果（制造商表述）。在本实验开展时，这是首款且唯一一款采用热通量技术的可穿戴体温计。
自2021年1月起，测量数据的可靠性显著提升，建议优先分析此日期之后的数据。2021年1月，原有设备因设计存在机械压力缺陷损坏，故更换为新设备。自2021年1月起，受试者/研究者自行设计了一款新型固定附件：采用魔术贴垂直粘贴于CORE设备本体，并环绕Polar Pro胸带形成固定环，以规避此前的机械缺陷。该设计带来了意外的附加效果：显著提升了数据可靠性，且CORE设备与皮肤的接触面积与贴合度均得到改善，因此测量结果更为可靠，可系统性反映昼夜节律特征。通过该新型固定附件，通常可将昼夜节律夜间时段简单定义为核心体温低于36.5℃的时段。
结合睡眠日记来看，该数据集是目前初步研究结果中最可靠的昼夜节律监测标志物之一。
关于数据列的说明：多数列均为加密乱码数据，仅核心体温、皮肤热通量1以及时间戳字段可被读取并用于分析，其余列均可且应当丢弃。具体而言，仅以下字段可被使用（源自制造商文档）：
* time [UTC-OFS=+0100]：数据点时间戳，格式为「DD:MM:YYYY hh:mm:ss」
* timestamp [us]：UNIX时间戳（单位为微秒，即自1970年1月1日起累计的微秒数）
* Hf_a0：热通路A的原始热通量传感器信号，以ADC计数为单位（无量纲）
* Temp_a0 [mC]：热通路A的未校正温度信号，单位为毫摄氏度（mC）
* Cbt [mC]：核心体温估计值，单位为毫摄氏度（mC）

需注意：将Cbt值除以1000即可得到以摄氏度为单位的核心体温数值。
以下字段无有效数据，均为随机填充值，可直接丢弃：hf_a1、Temp_a1、ax、ay、az、Battery_voltage [mV]。
热通量传感器A的电压可通过如下公式计算：
热通量传感器A电压（单位为微伏，μV）= hf_a0（单位为计数） × 1.953125 μV

传感器A的皮肤温度可通过如下公式计算：
皮肤温度A（单位为摄氏度，℃）= (temp_a0（单位为毫摄氏度，mC） – T0off) / 1000

其中T0off为偏移量，在每个CSV文件的文件头中仅定义一次。
2021年1月存在数据间隙：该时段为原有设备损坏至制造商更换新设备的过渡期。新设备采集的数据集始于2021年1月21日，搭配受试者自行设计的垂直魔术贴固定系统，该系统可防止设备损坏，并优化了皮肤贴合度，进而提升数据质量。建议优先分析2021年1月21日之后的数据，该时段数据噪声最低。
每个文件代表最长3.5天的单次采集会话，因此可通过拼接文件以研究更长时间跨度的节律变化。
额外说明：2021年5月29日T21:25:00，尝试对魔术贴布局进行优化，将垂直方向改为斜向，使CORE设备以菱形而非方形贴合皮肤，理论上可进一步提升皮肤贴合度与数据质量——此时胸带将在水平中线的两个边角施加压力，降低设备卡在垂直边角局部平衡状态的概率。但遗憾的是，该调整后数据质量反而下降（体温高低相位不再与实测睡眠-觉醒数据相关，且无明显高低相位出现），这表明设备的两个热通量传感器沿设备中线垂直分布（一个温度传感器位于顶部，另一个位于底部），因此尽管斜向布局的皮肤贴合度更高，垂直布局仍为最优选择。因此在2021年6月6日T01:00:00，魔术贴带的布局被重新改回垂直方向。斜向布局仅进行了一次采集会话，对应文件的命名亦体现了该调整。
数据后处理建议：如需对数据进行后处理以用于分析，需移除每个文件/新数据采集会话的前30分钟数据。这是因为传感器需要一定时间收敛至真实的当前体温：初始阶段传感器读数会偏高（因未佩戴时传感器处于低温状态，双热通量方程会输出偏高值），随后会逐渐降低直至达到平衡状态，此时的读数即可代表当前核心体温。

搜集汇总

背景与挑战

背景概述

该数据集记录了使用GreenTEG CORE Research设备采集的非24小时昼夜节律障碍患者的体温数据，采样频率为1Hz，包含核心体温和皮肤温度测量。数据质量在2021年1月设备更换后显著提升，建议优先使用此后数据进行分析。