EKG Classifier

Inspiration

Our inspiration for this project was one of our group member's mother having experience with cardiovascular care units and hearing about how much time is wasted of medical professional's time. Typically, three patient's EKGs are read by a medical professional, and we wanted to automate this process more and to combat workforce shortages. This will in turn lead to enhanced patient care and an automated process would help minimize human error when interpreting EKGs.

What it does

EKG Classifier will help automate the process to interpreting EKGs and has been trained to understand arrythmia versus a normal heartrate. We also sectioned the data between genders to see the differences for women's cardiovascular care in comparison to men. Generally, women are underrepresented in clinical trials and are more likely to be misdiagnosed for cardiac risk. Our program should detect problems quicker.

How we built it

Our project used an RNN model, LTSM (long term short memory). We implemented data cleaning to transform our data into a format suitable for modeling. We used data splitting which is splitting the dataset into training, validation, and test sets to evaluate model performance properly. The way we did LSTM-based ECG analysis using a public dataset of EKGS sorted by arrhythmia types.

Challenges we ran into

We had to take a long time with having the machine learning model train itself with data sets and due to the large quantity of data, it took hours to fully train the program. We had to redo it once which added a lot of time as well. In addition, we had to understand and learn the architecture of RNNs as well as understanding and using the datasets in raw waveform we acquired.

Accomplishments that we're proud of

We're proud for sticking to a project and working as hard as we did. We worked together as a team to deliver a result we can have some semblance of pride in.

What we learned

We learned that cardiovascular health is an underdeveloped field in medical technology. We also learned how to implement machine learning algorithms and LSTM. We learned to persevere in the face of adversity and extreme (EXTREME) fatigue.

What's next for Detecting EKG Irregularities with Machine Learning

We hope to see this concept be further developed so EKGs can become more accessible to a broader range of people. With simplified usage, the educational demand and technical complexities can be reduced so more economically disadvantaged areas can use EKGs with fewer trained medical professionals.