MDM (Medical Diagnostics Machine) (startup, overall, hw, b)

Picture of arduino testing with thermometer and heart rate monitor

Inspiration

We saw AI being applied in many fields, and We thought we could apply it to healthcare through personalized care in AI generated diagnoses. We saw the success of what Theranos could have been if they had a little less crazy in the equation, and thought a machine that makes suggestions for potential diagnoses could aid doctors and other healthcare professionals in their work.

What it does

The machine as it currently stands uses arduinos to carry out the simple tasks of qualitative measurement in the form of temperature and heart rate(both commonly taken by nurses routine checkups). For the qualitative measurements ie. symptoms, a series of questions are asked to the user to determine how they feel. The python script then inputs these into openAI's API and sends a request using the prompt "my temp is x and my heart rate is y bpm, heres some information about my current wellness state. What is your diagnosis?". The script then returns the API's response in the VS Code Terminal.

How we built it

We used arduino knowledge and schematics to create the heart rate monitor and thermometer. For the actual code, we had to plug the arduino jackets into a USB cable, so that it gets the data and is able to get the information from arduino (into the python script) as we had to decode arduino data in python script. To construct the thermometer and heart rate monitor, a thermal sensor and light sensor were used. The thermal sensor was quite simple: attach it to a 5V power source along with a 10kOHM resistor, and use arduino code to make the calculation that converts the thermal energy detection to temperature. The heart rate monitor uses a light sensor that detects when light traveling through your finger changes as your heart expands and contracts. This was done using a 5V power source, a 220 OHM resistor for the light source, and a 10kOHM resistor for the light sensor.

Challenges we ran into

For a good chunk of the project, the sensors were a bit finnicky and weren't functioning correctly. We just had to play around a bit and make sure all the connections were in order. Some of the calculations were off and we had to recualculate and make new formulas to account for variability in responses since the sensors were off sometimes. Additionally, we ran into compiling errors such as arduino ports not connected, so we had to revamp our code to account for this.

Accomplishments that we're proud of

We got it working, with pre-existing and symptoms from the patient, so its a very general AI generated diagnosis

What we learned

We learned that its always helpful to know how to transition between hardware and software, especially at their connection, because it took us a while to realize we could connect the arduino to a python script and just run it there