Join us during SPS annual meeting in Barcelona, Spain for our sponsored presentation with Amgen.
Leveraging Machine Learning for Automated ECG and Hemodynamic Analyses
Mr Joel Baublits, Senior Associate Scientist at Amgen
September 24, 2019 – 12:30-1:30 PM – Room 121
In recent years, cardiovascular safety scientists have adopted a number of analysis software platforms designed to streamline the evaluation of ECG and hemodynamic endpoints.
Post-processing with technologies such as pattern recognition can improve speed and accuracy of analysis using user-defined waveform libraries. While these technologies increase efficiency, the need to evaluate larger datasets (beat to beat), with increased depth (e.g. arrhythmia analysis), presents some conundrums: the need for significant level of human intervention, processing time, and most importantly, can be prone to high inter-user variability, a potential consequence of varying experience and analysis fatigue.
The goal of this study is to leverage historical data to guide a novel pattern recognition algorithm in performing automated ECG and hemodynamic analyses.
Analysis was conducted using a developer build of ecgAUTO (v184.108.40.206, emka TECHNOLOGIES, Paris, France).
Amgen historical data was used to construct master waveform libraries for multiple assays including conscious non-human primates and conscious/anesthetized beagle dogs. Model waveform selection was based on a composite mode, marking the P, QRS, and T “zones” independently using different waveforms from the master library.
Reference compound datasets, including known Ikr and Na channel blockers, were used to test the algorithm. New output data was compared with historical data to determine performance.
The algorithm delivered high fidelity data that correlated well with historical analyses, suggesting that this new approach will be able to leverage historical data in a dynamic manner that can scale with increasing data volume. In addition, decreasing manual overreading will result in greater efficiency, delivering high quality data that is less prone to inter-user variability.
This session is a Sponsored Presentation. Although not an official part of the SPS Annual Meeting scientific program, its presentation is permitted by the Society
How to find us at SPS?
Safety Pharmacology 2019 Annual Meeting
September 23- September 26, 2019