LLNL

Location

Livermore
Year

2024
Stack Use

Numpy

Pandas

Scikit-Learn

PyTorch
Share

Exciting Problem in Cardiology -- Machine Learning Diagnosis of Irregular Heartbeats

This summer, I had the immense honor of participating in the Data Science Challenge at Livermore. We were introduced to a heart disease problem and tasked with applying machine learning to tackle a series of challenges.

On the third day of the event, I had the privilege of visiting Discovery Center LLNL. The featured exhibitions included the NIF Target Chamber, inventory management, and energy and climate science. At the Discover Center, we had the opportunity to learn more and even experienced hands-on activities like generating power through human effort to drive intriguing devices.

Throughout the event, our team collaborated closely, tackling challenges and solving problems together. It was during this event that I learned and built my first neural network models, an incredibly valuable experience.

Approaching to end of program, we created a poster showcasing our project results. On the final day, we proudly presented and explained our poster to the guests, successfully completing our task. This summer, I gained invaluable experience and earned a completion certificate.

In the end, I would like to extend my heartfelt thanks to all my team members, mentors, and the supportive staff!

Some of the valuable skills I took away from this experience include:

Abstract & Background

Arrhythmia is a medical condition characterized by irregular heartbeats, which, if left untreated, can have life-threatening effects. This project leverages machine learning to detect heart abnormalities using inexpensive electrocardiogram (ECG) data. Our model can classify ECG data as usual or one of four abnormal heartbeats with a false negative rate of only 0.87%. Additionally, our model's predictions for myocardial activation times are, on average, only 2 ms off. These results suggest that machine learning is a promising, cost-effective solution for arrhythmia detection.

Heart disease is the leading cause of death in the United States. Arrhythmias are a result of underlying heart problems and classifying them can provide life-saving care. A common method of getting heartbeat data is a standard 12 lead ECG, where electric signals from the heart are measured. We aim to use these signals to predict when specific parts of the heart activate.

Lawrence Livermore
National Laboratory

Exciting Problem in Cardiology -- Machine Learning Diagnosis of Irregular Heartbeats

Abstract & Background

Poster