PetWellTech develops hardware and software for monitoring pet health and associated activities.
The goal of the pilot project is to investigate possibilities of monitoring VOCs in dog breath with hardware available on the market and build a prototype solution that can detect for example acetone in dog breath.
Image | Sensor Name | Description |
---|---|---|
"Electronic nose", Bosch BME688 sensor | A development kit consisting of 8 individual BME688 sensors was used to collect data along with a Thing Plus - ESP32 WROOM (Micro-B). Documentation for the BME688 sensor can be found here. | |
SparkFun Atmospheric Sensor Breakout - BME280 | Getting started guide | |
Adafruit CCS811 Air Quality Sensor | Documentation |
Controlled environment in a plastic box. Volume of the box is 0,013 cubic meters. The amount of acetone injected was 0.01 ml
- Density of acetone: 0.791 g/ml.
- Molecular weight of acetone: 58.08 g/mol.
- Temperature: 35°C (converted to Kelvin: 308.15 K).
- Gas constant: 8.3145 J/(mol⋅K).
-
Convert Liquid Volume to Mass:
- Mass of acetone = 0.01 ml x 0.791 g/ml = 0.00791 g
-
Convert Mass to Moles:
- Moles of acetone = mass / molecular weight = 0.00791 g / 58.08 g/mol ≈ 0.000136 mol
-
Use Ideal Gas Law to Find Volume of Gas at 35°C:
- Calculate volume using V = nRT / p, assuming standard pressure (101325 Pa). For example using this calculator for ideal gas law.
- Volume ≈ 3.44 ml
-
Convert to ppm:
- Concentration (ppm) = (Volume of gas / Volume of box) x 1,000,000
- Concentration ≈ (3.44 ml / 13000 ml) x 1,000,000 ≈ 265 ppm
The acetone was added to the enclosed box using a small syringe at the beginning of the test. There was no more acetone added during the test.
![Box](images/syringe.JPG) ![Box](images/aceton.JPG)The table below shows the data collected for the different acetone concentrations and for clean air, as well as the duration of the data collection.
*Temperature (°C) | *Humidity (%) | *Pressure (hPa) | *Gas Resistance (Ω) | Duration (mins) | |
---|---|---|---|---|---|
Air training | 3.501277e+01 | 1.925969e+01 | 1.020081e+03 | 3.489649e+07 | 45 |
Air test | 3.459404e+01 | 1.943559e+01 | 1.018670e+03 | 3.361244e+07 | 45 |
Acetone training | 3.635105e+01 | 2.050594e+01 | 1.021596e+03 | 1.580791e+07 | 15 |
Acetone test | 3.543839e+01 | 1.962744e+01 | 1.018924e+03 | 1.418220e+07 | 15 |
Temperature (°C) | Humidity (%) | Pressure (hPa) | Gas Resistance (MΩ) | Duration (mins) | |
---|---|---|---|---|---|
Air training | 35.01 | 19.26 | 1020.08 | 34.90 | 45 |
Air test | 34.59 | 19.44 | 1018.67 | 33.61 | 45 |
Acetone training | 36.35 | 20.51 | 1021.60 | 15.81 | 15 |
Acetone test | 35.44 | 19.63 | 1018.92 | 14.18 | 15 |
The image below shows the resistance change over time when the sensor was exposed to acetone and air. The highlighted area shows the resistance change when the sensor was exposed to acetone for the first cycle. WWhen the sensor was exposed to air the resistance decreased notably. The average resistance here was X
Compared to the reference air, the resistance change for acetone was significant. This is a good indication that the sensor can be trained to detect it at least at relatively high concentration.
Using the BME studio software an algorithm was trained to detect acetone. The training data set was used to train the data with a 70/30 split. The image below shows the confusion matrix for the training data set.
The table below shows the accuracy of the training data set.
Accuracy | F1 score | False positive |
---|---|---|
96.01% | 96.11% | 4.01% |
USing the BME studio software the trained algorithm was tested with the test data set. The image below shows the confusion matrix for the test data set. This show that the algorithm was able to detect acetone with a high accuracy.
Accuracy | F1 score | False positive |
---|---|---|
91.54% | 91.53% | 8.55% |
- Z. Wang, C. Wang and P. Lathan, Breath Acetone Analysis of Diabetic Dogs Using a Cavity Ringdown Breath Analyzer, in IEEE Sensors Journal, doi: 10.1109/JSEN.2013.2293705.
- Saasa V, Malwela T, Beukes M, Mokgotho M, Liu CP, Mwakikunga B. Sensing Technologies for Detection of Acetone in Human Breath for Diabetes Diagnosis and Monitoring. Diagnostics (Basel). 2018 Jan 31;8(1):12. doi: 10.3390/diagnostics8010012.