Detecting Tremors Using Leap Motion
The inspiration came from the recent development of therapeutic supports for physicians. One key example that come to the teams mind the need for development in being able to accurately diagnose degenerative disorders. From that, given some past knowledge two of the most major types of tumours are essential tremor (ET) and the tremor of Parkinson’s disease (PD). _ A tremor is defined as an involuntary, rhythmic, and roughly sinusoidal movement of one or more body parts._ After being able to see some of the hardware available, we realized that we could use the Leap Motion - a Human-Machine-Interface that provides a method of position tracking with extremely precise accuracy, and high throughput rate of up to 300 samples per second. From that our idea was born - Tremble.
What it does
Tremble is a diagnostic tool for physicians that allow them to more accurately detect the presence of tremors in their patients in real time providing extremely accurate data about their patients.
How we built it
We built Tremble using mainly Python utilizing elements of the Leap Motion library. Additionally we used Python to develop the FFT algorithm we implemented. From there, we used Typescript and the Angular 7 framework to develop the front-end interface that the physicians will interact with.
Challenges we ran into
The main challenges we ran through was implementing the many parts of the project and bring them all together. We had difficulty connecting our backend Leap Motion framework with our algorithm and finally getting that on to our webpage. Along with that, we were using developing an implementation of an algorithm of Fast Fourier transformation to change a time series distribution to a frequency series and it took along time to understand the fundamental math behind this algorithm and apply it to our solution.
Accomplishments that we're proud of
My team and I were able to come in and build something that can actually have real value in the real world. The proof of concept my team created actually works and scaling this is definitely possible.
What we learned
My team and I struggled a lot but that resulting in a lot of learning. Our very first step we were trying to get the base Leap Motion package to work and it wouldn't work for 4 hours until we all realized the it was some path issue that none of us knew about before. On top of that, we learned about new algorithms such as the FFT and how to best optimize a website for physicians.
What's next for Tremble
We see great opportunity in the future development of Tremble. As of now this is a very untouched industry. Once there is more data we want to be able build a machine learning model to be able to train a model to recognize different types of hand tremors and associate them with different types of tremor associated diseases along with being able to detect the severity of these diseases.