SenseAir | Devpost

Inspiration

Respiratory infections like pneumonia remain difficult to detect quickly and non-invasively. This inspired us to explore breath-based detection methods. We were particularly interested in whether VOC patterns associated with the bacteria Streptococcus pneumoniae could be classified using a compact gas sensor and machine learning.

What it does

SenseAir uses responses from the BME688 to detect patterns in volatile organic compounds that resemble pneumonia-associated breath signatures. Instead of identifying specific chemicals, it classifies overall VOC response patterns into baseline air, single-compound exposure, or pneumonia-like mixtures.

How we built it

After researching the VOCs of the chosen bacteria, we generated a realistic synthetic dataset simulating MOX sensor resistance changes under controlled VOC exposure scenarios, incorporating environmental variables like humidity and temperature.

Challenges we ran into

A large issue we ran into was in implementing the live data stream from our Uno Q on our backend to our front-end. We eventually decided to forgo using a WebSocket because of this.

Another challenge was 3D printing the case for our hardware. Between failed prints, parts not aligning, and warping mouthpieces, eventually, the final case was produced.

The main challenge that we spent the majority of our time on was training our machine learning model to have an accurate reading. Our first few models yielded very low accuracies. Through switching our model to a decision tree, implementing confusion matrices, and getting more data, we eventually were left with the model we have now.

Accomplishments that we're proud of

We are proud to have built a balanced dataset that models realistic gas sensor responses to pneumonia-like VOC mixtures. We developed a clear framing for pattern-based infection screening rather than direct bacterial detection.

What we learned

For all of us, this was the first time we had an air sensor in our hands, which was a big learning curve. While some of the members on our team are studying in the Department of Bioengineering here at SCU, they have grown their knowledge of infectious diseases and their associated biomarkers extensively.

In a similar vein, we have never created our own custom design and brought it to life using 3D-Printing. For our BIOE, who dedicated the night to testing prints and refining over and over, we are very grateful.