Clinical Insights

Investigational Use Only

This document is intended for (data) scientists who are very familiar with the properties of sample-level (raw) physiologic and behavioral data as can be captured with Byteflies Kits.

When in doubt, do not attempt to follow these instructions and contact us.

Table of contents

Clinical Insights

Introduction

The Byteflies Cloud processes the sample-level (raw) data recorded by Sensor Dots into certain Clinical Insights, i.e. vital signs and other clinically useful information. This document describes which Clinical Insights are generated and how to leverage them for your own work.

Clinical Insights that underwent rigorous clinical evaluation are clearly labeled with a icon. These files can be safely used under the Byteflies Kit indications for use, and details on how they are calculated are discussed in this document.

The following Clinical Insights are available:

Sample-level Signal	Clinical Insights	Details
ECG	`HEARTRATE` `EDR_RESP`	Heart Rate (HR) as derived from an ECG signal Respiratory Rate (RR) as derived from an ECG signal
ACC	`ACTINDEX`	Activity Index (AI) as derived from an ACC signal

Unless otherwise indicated, all content of this document should be considered strictly for investigational use only.

Additional Clinical Insights are available that are “For investigational use only”. Unless Byteflies explicitly instructed you to use any of the investigational use files, you will not need these!

All Clinical Insights are specific to a Signal and are automatically calculated as soon as a Recording is uploaded to the Byteflies Cloud, provided that Recording is at least 1 minute or longer.

Practical Tips

Conversion Factor and Signal Units

Conversion factors are multipliers that convert sample-level data into commonly used international standard units for the signal type. Conversion factors for a specific Signal can be found via the API.

For a summary of the resulting units, refer to the API documentation.

File Structure

The files that are attached to a Recording record are all in CSV (comma separated values) format.

Sample-level files have the following structure:

time	channel
Time in seconds relative to the start of the recording (float)	Raw signal value (integer)

Note that for triaxial data, such as for ACC and GYR files, 3 channels are available in a single file, labeled channel1 (X-axis), channel2 (Y-axis), and channel3 (Z-axis).

To convert the time column to Universal Coordinated Time (UTC), add the UNIX timestamp that you can access programmatically over the API. It may be necessary to adjust for your local timezone!

Clinical Insight files typically have the following structure:

Time(s)	Insight(units)
Time in seconds relative to the start of the recording (float)	Insight value (integer or float)

Note that this structure is not followed by all files. Refer to the information for a specific Clinical Insight below.

Accelerometer Data

Sample-level accelerometer (ACC) data measures acceleration in 3 dimensions. Depending on the location of the Sensor Dot, it can provide granular information on a patient’s movements and behavior.

Activity Index (AI)

CE mark

The Activity Index (AI) is a measure of total activity level derived from ACC data. It has been calibrated to yield zero activity when the Sensor Dot is completely still (e.g. on a table).

Label: ACTINDEX
Units: g (g-force)
Output: Mean AI value for every 10 sec window.

Gyroscope Data

Sample-level gyroscope (GYR) data measures rotational velocity in 3 dimensions. Depending on the location of the Sensor Dot, it can provide granular information on a patient’s movements and behavior.

Electrocardiography Data

Sample-level electrocardiography (ECG) data measures the biopotential signal generated by the conductive system of the heart.

R-peaks

R-peaks are a specific feature of an ECG signal that denote the depolarization of the cardiac ventricles.

Label: RPEAK
Units: Index: Position of an R-peak relative to the sample number in the Signal or Time(s): Position relative to the start of the Recording in seconds.

Heart Rate (HR)

CE mark

From the R-peak positions, the heart rate (HR) in beats-per-minute (BPM) can be calculated.

Label: HEARTRATE
Units: BPM (beats-per-minute)
Output: Median HR value for every 10 sec window.

ECG-derived Respiration (EDR)

The ECG Signal also contains respiratory respiration as a consequence of chest wall motion and chest impedance changes with changing lung air volume. The ECG-derived Respiration (EDR) is a proxy Signal derived from ECG which represents respiratory modulation.

Label: EDR
Units: A new representation of an ECG Signal that amplifies respiratory modulation.

The EDR Signal should only be used when the AI is below 0.25 g. Otherwise, the results cannot be trusted.

EDR-peaks

The EDR-peaks are calculated to identify the respiratory cycles represented in the EDR Signal.

Label: EDR_PEAK
Units: Index: Position of an EDR-peak relative to the sample number in the Signal or Time(s): Position relative to the start of the Recording in seconds.

Respiratory Rate (RR)

CE mark

From the EDR-peak position, the respiratory rate (RR) in breaths-per-minute (BrPM) can be calculated.

Label: EDR_RESP
Units: BrPM (breaths-per-minute)
Output: Median RR value for every 60 sec window.

ECG Artifacts

In order to judge the quality of an ECG Signal, specific signal artifacts are detected. Based on these signal artifacts, a quality score is calculated. This quality score is visualized to the user in one of two ways via an API call which returns a PASS, CHECK, or FAIL label.

Label: ECG_ARTFCTS
Artifacts: OUTLIER An ECG event, such as an R-peak that is in an unexpected position or GAP a pause in R-peak detection that is too long to be physiologic, i.e. most likely caused by a loss of Signal.
Output: A start_index and end_index that denotes the relative position of the artifact to the sample number in the Signal, and a artifact label.

Electroencephalography Data

Sample-level electroencephalography (EEG) data measures the biopotential signal generated by the brain.

EEG Artifacts

In order to judge the quality of an EEG Signal, specific signal artifacts are detected. Based on these signal artifacts, a quality score is calculated. This quality score is visualized to the user via an API call which returns a PASS, CHECK, or FAIL label.

Label: EEG_ARTFCTS
Artifacts: MINOR A transient low amplitude EEG artifact, MAJOR A sustained high amplitude EEG artifact, NO_SIGNAL a loss of signal, or FREQ_ANOMALY an abnormality on the EEG frequency spectrum.
Output: A start_index and end_index that denotes the relative position of the artifact to the sample number in the Signal, and a artifact label.

Electromyography Data

Sample-level electromyography (EMG) data measures the biopotential signal generated by a skeletal muscle.

Electrooculography Data

Sample-level electrooculography (EOG) data measures the biopotential signal across the eye(s) to measure eye movement.

Lead-off Data

Lead-off is a diagnostic signal that indicates if the Sensor Dot is receiving a measurable biopotential input signal. If Lead-off = 1, no proper input signal is detected. If Lead-off = 0 a measurable input signal is detected. The Lead-off file stores a 1 or 0 value with a timestamp every time a change is detected. For instance, if the Sensor Dot is not properly connected to a cradle or biopotential electrodes, as specified in the instructions for use, Lead-off will be set to 1.

Event Log

The Event Log is a diagnostic signal that collects all the events that occur during a recording. The file is a JSON encoded array of events:

[{
  "Type": "StartOfStream",
  "Time": 0
},
{
  "Type": "EndOfStream",
  "Time": 2623.017493896
}
]

The Time-field has as unit seconds from the start of a recording. A non-exhaustive list of possible Types:

StartOfRecording: start of the recording
EndOfRecording: end of a recording
ExternalReset: external reset is triggered

Battery Level

The Battery Level is a diagnostic signal that represents the battery drain of the Sensor Dot during a Recording.