MedMeasure

Inspiration

There was no particular motivation. It's just a project that I think would be useful in busy healthcare places where the priorities are not always on small, but can be dangerous injuries such as wounds.

What it does

MedMeasure is a camera-vision-based project. It scans a picture of a wound alongside a reference box. It then gives the user the dimensions and area of the wound. This can then be recorded, saved and made into a graph which the user can download to track the healing progress. You can also connect to an Arduino to build hardware projects where this program is more useful.

How we built it

We first plan what the project requires a computer-vision program with image recognition and a way to record the data. We decided to use Python for this project since it has every library we need. We integrated Codex to accelerate some of the coding process and fix some bugs. We repeat this process for each step of the project which divides into 5 main step. First step is to initialize the computer vision system. The second is to calibrate and make reference points available with every orientation of the reference box. After that, we need to calibrate the wound detection which only scans for red and can lead to scanning the whole limbs/area. Then, we add a program that allows Arduinos to connect to the system for real time data acquisition. Finally, we fix bugs and add functions that are little but make the user interface a lot easier to use.

Challenges we ran into

The biggest problem that we ran into was the calibration step. We have to find a way to orient the reference box so that it's easy for the program to read. This includes folding, stretching and spinning the pictures around. Eventually, we made a code that do that within the app.

Accomplishments that we're proud of

We are proud to have finished a prototype and get Codex to integrate a code that let you connect to an Arduino.

What we learned

We learned more about computer vision and how to translate what we see to pixels and matrices that a program can see and understand. We also learned how to calibrate and orient reference points to a particular orientation.

What's next for MedMeasure

Right now, it's a prototype. We want to test more with a broad dataset. We hope that we can integrate Machine Learning into this program to accelerate calibration and data validation. We want this to work well with every skin color as this prototype struggles with darker skin color because the program is not great with shadows, yet.