The Office of Informatics recently added electrocardiography (ECG) waveform data to augment the existing ECG discrete elements already available through i2b2.
An ECG measures changes in the electrical activity of the heart over time, and is the most commonly collected data in cardiovascular research. While relatively easy to measure, ECG waveforms reflect a wealth of information on underlying aspects of heart disease. ECG discrete elements (such as the PR interval, QRS interval/duration, QT interval and RR interval) are already being used by our researchers. Now 3,740,036 historical waveforms, extracted with support from Dr. Oguz Akbilgic, have been added to the Translational Data Warehouse. Each waveform has 12 leads and each lead has its own encoded data which can be extracted into an excel file for use in research, particularly in machine learning based research. In addition to the historical waveforms now available, ongoing prospective extraction continues for all current ECG tests from Muse (Wake’s cardiology information system).
“Cardiovascular disease has consistently been the leading cause of death worldwide. Screening for cardiovascular disease, early diagnosis and intervention is the key to reduce the burden of cardiovascular disease treatment in healthcare system while improving patient care. Yet, most gold standard diagnosis protocols require high level imaging such as cardiac MRI, CT, and ECHO. These imaging modalities are not suitable for screening large patient populations due to associated cost, low accessibility and availability across the nation. Electrocardiogram (ECG), on the other hand, is a low cost, widely available, and now is even embedded into consumer level devices such as smartwatches. There is a growing literature showing the power of raw time-to-voltage ECG data to detect and predict risk for several major cardiovascular and non-cardiovascular diseases, when analyzed by advanced artificial intelligence methods. The critical barrier in developing such ECG-AI disease detection and prediction models is that such raw ECG data is typically stored in separate servers (e.g. cardiology information systems) in encrypted fashion, and never makes it to the EHR. At Wake Forest University School of Medicine, we have removed this critical barrier by exporting and decrypting close to 4 million retrospective ECGs, and creating a pipeline to automatically export prospective ECGs to our research database. Now, this ECG repository is linkable to research data in i2b2 to enable numerous research and quality improvement projects.”– Dr. Oguz Akbilgic, Cardiovascular Medicine, Biomedical Informatics