Safe automatic screening for AF using Cardiolund ECG Parser

There is a growing interest in identifying AF patients at an early stage to be able to intervene the progression of the condition and to prevent stroke cases. Identifying AF patients may be complicated by factors such as AF being without symptoms (silent AF) for many patients, and the variety of non-critical arrhythmia that are found in ECGs in an ageing population.

A simple algorithm for disguising between normal sinus rhythm (SR) and AF will not be very helpful. Automated screening for AF in an older population requires an algorithm that can distinguish between AF and multiple other arrhythmia.

The Cardiolund ECG Parser has been evaluated in large scale clinical study on screening for atrial fibrillation (the STROKESTOP study).

The study involved 3209 individuals (age 79-80) and a total of 80 149 intermittent electrocardiogram (ECG) recordings performed using Zenicor single-lead thumb-ECG recorders. The participants recorded a 30 sec ECG twice per day for 2 weeks using the device unassisted in their own home. Persons with known AF were excluded from the study. The results of manual interpretation (gold standard) was compared to the ECG Parser.

The study concludes:

  • The ECG Parser is very well-suited for the purpose of AF screening, as it safely identifies normal ECGs and as it reduces the need for manual interpretation by >85%.
  • The sensitivity for AF detection by the algorithm was 97.8% (95% CI 95.4 – 99.1).
  • The total specificity of the algorithm was 88.2% (95% CI 88.0 – 88.4).
  • If using the algorithm for screening, as alternative to manual review of each recording, the cost reduction would (depending on salary) amount to approx. 12 570 Euros/1000 screened.

Further details are available in the full text of the paper:

Emma Svennberg, Martin Stridh, Johan Engdahl, Faris Al-Khalili, Leif Friberg, Viveka Frykman, Mårten Rosenqvist; Safe automatic one-lead electrocardiogram analysis in screening for atrial fibrillation, EP Europace, Volume 19, Issue 9, 1 September 2017, Pages 1449–1453,

In summary, the Cardiolund ECG Parser safely identifies normal ECGs and reduces the need for manual interpretation of individual ECGs with 85%. The sensitivity of the system in identifying AF is 100% on an individual level.