Real-Time Extraction and Analysis of Key Morphological Features in the Electrocardiogram, for Data Compression and Clinical Decision Support
Name
AnkitAAAI2011.pdf
Size
95.41 KB
Format
Adobe PDF
Checksum (MD5)
7cd32d135c3fa6013b4cf1a6c0371d7e
Author(s)
Gordhandas, Ankit
Heldt, Thomas
Verghese, George C.
Date Issued
March 2011
Journal
Computational Physiology: Papers from the AAAI 2011 Spring Symposium
Publisher
American Association for Artificial Intelligence
Citation
Gordhandas, Ankit J., Thomas Heldt and George C. Verghese. "Real-Time Extraction and Analysis of Key Morphological Features in the Electrocardiogram, for Data Compression and Clinical Decision Support." In Computational physiology: papers from the AAAI spring symposium, March 21 - 23, 2011, Stanford University, Stanford, California USA. Copyright c 2011, Association for the Advancement of Artificial Intelligence.
Version
Final published version
Abstract
Massive amounts of clinical data can now be collected by stand-alone or wearable monitors over extended periods of time. One key challenge is to convert the volumes of raw data into clinically relevant and actionable information, ideally in real-time. This becomes imperative
especially in the domain of wearable monitors, where power and memory constraints prevent continuous communication of raw, uncompressed data to a base station for a health care provider. We focus here on algorithmic approaches to extract clinically meaningful
information from the electrocardiogram (ECG) in realtime. We use a curve-length transform to identify, and aggregate from beat to beat, physiologically relevant timing information, such as the onsets and offsets of P-waves, QRS complexes, and T-waves, along with their respective
magnitudes. Each heartbeat is thus parametrized in terms of 12 variables. Assuming a nominal heart-rate of 70 beats per minute, and a sampling frequency of 250
Hz, each beat has approximately 215 samples. Reducing each beat to 12 samples thus gives an 18-fold compression. An exponentially-weighted sliding average of the identified
morphological features over the preceding twenty beats is also stored. Whenever any feature deviates significantly from its stored weighted average, the algorithm registers an alarm and also retains the raw ECG data of the 5 beats immediately preceding and following the anomalous occurrence, for a later review by a clinician.
MIT Department
Massachusetts Institute of Technology. Department of Electrical Engineering and Computer Science
Massachusetts Institute of Technology. Research Laboratory of Electronics
Terms of Use
Article is made available in accordance with the publisher's policy and may be subject to US copyright law. Please refer to the publisher's site for terms of use.
Persistent DSpace Link
DOI of Published Version
http://www.aaai.org/ocs/index.php/SSS/SSS11/paper/view/2493/2904
