Real-Time Detection and Classification of Heart Arrhythmia Using ECG Feature Detection

The Problem

Studying how long-term vaping affects the heart means reading the waveform in enough detail to see the beat structure, and a clinical twelve-lead machine is out of reach on a student research budget. A two-student engineering thesis group needed a low-cost rig that could take a trace and tell them what it showed.

What We Built

We built a single-lead ECG rig and the desktop software that reads it. An Arduino front end streams the raw signal and goes silent the moment an electrode comes loose, so a detached lead produces no trace rather than a convincing bad one. The application records a ten-second window, delineates the P, Q, R, S, and T waves, computes heart rate, peak amplitudes, and eight clinical intervals with their variability, and flags atrial fibrillation, atrial flutter, and sinus tachycardia against the thresholds the study specified. Researchers can reload an earlier session, trim a noisy head or tail, recompute, and read every fiducial point off the plot. It is research instrumentation, not a medical device, and nothing it reports is validated for diagnosis. Turning a noisy biological signal into measurements a researcher can defend is the hard part, and that is the part we built.

Results

• 3 arrhythmias

  atrial fibrillation, atrial flutter and sinus tachycardia, flagged against the study's own thresholds

• 8 intervals

  clinical intervals and their variability computed from every ten-second recording

• P through T

  all five waveform points delineated, with the P, R and T onsets and offsets marked on the plot

• Fails loud

  a detached electrode produces no trace at all, never a plausible-looking bad one

Techstack