ECG Parser version 2.5 is released

Cardiolund have today released ECG Parser version 2.5. This is a performance improving release (no functional changes) focussing on the following two areas:

1. Signal categorisation and tagging for (thumb-ECG and similar) signals analysed using the NotAttached lead config. The changes results in significant performance enhancements in the use of categories and tags for AF detection, e.g., when used in a screening process of short thumb ECG recordings.

2. Verification of VES detections in (Holter, 12-leads, etc.) signals analysed using other the NotAttached lead config. The changes results in significant improvement in the VES detection PPV (positive predictive value).

Highlights for release v2.5:

• Improved handling of large amplitude VES in single lead thumb-ECG, and similar, recordings analysed using lead configuration NotAttached (thumb-ECG and similar).
• New AI/ML models for detecting bad and similar beats, used for improvement of categorisation/tagging of noisy signals.
• Reduction in the number of non-AF signals in category 2 (Irregular Sequence), measured on the STROKESTOP database, is 21% when compared to release v2.4, 38% when compares to release v2.3, and 39% when compared to release v.2.2.2.
• New AI model for validating VES detections in Holter, 12-lead, and similar types of recordings.
• Increase in VES PPV on Holter databases to about 92-95% (more than 6% increase compared to previous releases), and VES sensitivity has increased to above 90 on all test databases.
• Expanded performance testing related to requirements in international standards using two additional test databases.

Read about the ECG Parser on the product page, or contact us for more information.