# Breathing Rate

## Description 

Breathing rate (BR) corresponds to the number of times the user inhales and exhales, expressed as a rate per minute (Range: up to 50 breaths/minute). 12-25 breaths/minute is considered healthy/normal.

## Method 

BR measured by BitDoctor was assessed against BR as measured with a respiration belt. All tests were conducted by researchers with experience in data collection, signal processing, and machine learning. 

## Participants 

The subjects were recruited from Affiliated Hospitals. There were no exclusion criteria when recruiting participants, although most participants were healthy adults above 18 years of age. Table 9 displays a summary of participant characteristics.

| Participants              | Composition          |
| ------------------------- | -------------------- |
| Average Age               | 41.2                 |
| Standard deviation of age | 15.2                 |
| Gender distribution       | 48% Male; 52% Female |

                                                 *Table 9: Characteristics of participants for breathing*

## Additional Data Collection Setup 

A mechanical respiration belt was attached around the thoracic region of each participant and then connected to a BIOPAC recording device. Figure 12 depicts the data collection setup.

<figure><img src="https://101746475-files.gitbook.io/~/files/v0/b/gitbook-x-prod.appspot.com/o/spaces%2FbQD4HPZ4Ef0SW88Edu1U%2Fuploads%2Ffqn65lbsqZXr1ye5mJPO%2Fd54.jpg?alt=media&#x26;token=17b307f2-9bf9-4ce1-8732-114d039d26e8" alt="" width="347"><figcaption><p>      <em>Figure 12: BR test setup</em></p></figcaption></figure>

## Additional Data Collection Procedure

Oscillations in the tension of the respiration belt (capturing inspiration and expiration) and video recordings of the participant’s face were captured in a 2-minute recording.

## Analysis of Accuracy & Reliability&#x20;

Reference BR was calculated by counting the number of breathing cycles (inspiration + expiration) captured by the respiration belt, and then expressing this value as a per-minute rate.&#x20;

BitDoctor-based BR was calculated by using the breathing detection algorithm to predict the number of breaths based on facial blood flow information, and then expressing this value as a per-minute rate.&#x20;

Breathing rate accuracy was then expressed as a Pearson correlation between participants’ reference breathing rate (as determined by respiration belt/BIOPAC) and breathing rate estimated by BitDoctor.

## Results&#x20;

The accuracy and reliability of BitDoctor BR is summarized in Table 10.

| Physiological Measure | Accuracy (%) | Reliability (%) |
| :-------------------: | :----------: | :-------------: |
|       Breathing       |      >90     |        99       |

&#x20;                                                 *Table 10: Accuracy and reliability of BitDoctor BR.*

The Pearson correlation between participants’ reference BR (as determined via respiration belt/BIOPAC) and BR estimated by BitDoctor was r=0.9. Figure 13 depicts the relation between reference BR (ground truth) and BitDoctor BR (predicted value)

<figure><img src="https://101746475-files.gitbook.io/~/files/v0/b/gitbook-x-prod.appspot.com/o/spaces%2FbQD4HPZ4Ef0SW88Edu1U%2Fuploads%2FyVipMxhPTm1ZB1uX7Gd7%2Fd53.jpg?alt=media&#x26;token=690e0d7d-895a-4172-b554-feee377fc32d" alt="" width="563"><figcaption></figcaption></figure>

*Figure 13: Scatter plot depicting the relation between reference (ground truth) BR measured with a respiration belt and BR estimated from BitDoctor. The equation characterizes the line of best fit.*