We have built a powerful software tool we call the ECG Parser. This software transforms an ECG signal, or an entire database of ECG signals, into a very detailed description of the content.
The derived ECG description is made available in a machine-readable format, convenient for use in diagnostics, feature analysis, statistics, visualisation, and a range of other purposes.
One such very promising purpose is data mining, for which the ECG description is very apt. You can create algorithms for automatic search of large ECG databases for different types of events, or feature patterns.
There are some fundamental differences between conventional ECG analysis software, and Cardiolund’s ECG parser.
The objective of a traditional ECG analysis software is to provide a diagnosis, or to present findings that support the physician in diagnosing the patient.
The objective of our ECG parser is to deliver a detailed technical description of the entire content of the ECG signal, in a format that allows for other tools to make use of the description for many diverse purposes, including diagnosis.
Obviously, all ECG analysis software are based on some form of ECG parsing, and automatic diagnostics is a natural add-on to an efficient ECG parser. And, it is very likely that the better the ECG parsing, the better the diagnostic performance.
The advantage of basing automatic diagnostics on a detailed description of the ECG, as resulting from the parsing process, is that there will be clear connections between signal features and diagnoses. Any decision based on the ECG Parser description is associated with a detailed set of evidence, which can also be made available to the user.
Unlike most traditional ECG software, the ECG Parser is available as a completely hardware-decoupled software product. It is therefore possible to employ the software to keep large ECG databases alive far beyond the lifetime, and other constraining limitations, of the recording system.
This opens for exciting new possibilities including: Searching in large ECG databases, signal categorisation for quick identification of an important condition, comparison between specific cases and groups, and application of statistical approaches (machine learning techniques) to gain new insights from an ECG database.
Searching in ECG data is a stirring, but complex, field that span possibilities ranging from searching for irregular beat patterns, or a large number of early beats per minute, to searching for patients with a certain development over time in yearly recordings, or searching for patients with the same type of response to an anti-arrhythmic drug.
With this technology you can answer very detailed analysis tasks, which cannot be performed manually, such as: Count and calculate trends for the number of early beats over time for patients that later developed bigemini or atrial fibrillation.
The thorough analysis done by Cardiolund’s ECG Parser, with very detailed description of types of content and features, takes searching of large databases to a completely new level, there are now a seemingly endless opportunity to learn from existing ECG databases.
Earlier ECG systems could search for AF diagnosis, or large RR variability, but with a new generation of ECG parsers, ECG systems can search for specific beat patterns, specific morphologies, and specific trends between recordings, or internally within recordings.
Categorisation of ECG data is essential in situations where ECG databases grow beyond what is possible to review manually. Home-based simplified ECG measurements for screening of atrial fibrillation is one such area, where ECG parsers are used to separate cases for review from cases with no motivation to review.
The Cardiolund ECG Parser categorises signals into twelve distinct categories that can be used for guiding the efforts of the clinical staff to focus on the cases they are looking for. With the current algorithms, this automatic categorisation of ECG data reduces the workload for reviewing data by a factor of 10. A claim that has been validated in a scientific study.
It is beyond doubt that clinical staff all over the world will soon rely on having parsed ECG databases to guide them to the important cases that needs treatment, and providing near unlimited new information for clinical researchers.
You can apply ECG parsing with great benefit to data from new devices, but equally important you can also apply ECG parsing to large ECG databases that exists in hospitals, research centres and companies, giving life and increasing the value of databases that otherwise would not provide more value.
If you want to know more about Cardiolund Research and the future of ECG analysis, don’t hesitate to contact us.