pyCardiograms

Python DSP & LSTM Waveform Prediction Library

Organization: Tensr.AI

Key Result & Impact Benchmark

Automated cardiac cycle fiducial point extraction with high clinical timing fidelity.

Technical Overview

pyCardiograms is the core Python data science library engineered at Tensr.AI to process raw biomedical time-series data captured by the Teresa Band. The package implements digital signal processing filters (Butterworth bandpass, wavelet denoising) to isolate mechanical cardiac vibrations from motion artifacts. It features a trained LSTM recurrent neural network that automatically annotates AO (Aortic Opening), AC (Aortic Closure), MO (Mitral Opening), and MC (Mitral Closure) fiducial points across SCG, ECG, and PPG waveforms.

Key Technical Highlights

Engineered automated signal denoising pipelines handling baseline wander and muscle motion artifacts.
Trained LSTM models for precise fiducial peak location across multimodal cardiac cycles.
Scaled R&D team to 15 engineers and data scientists at Tensr.AI.

Technologies & Frameworks

Python

LSTM Networks

SciPy DSP

PyTorch

ECG / PPG / SCG Annotation