Clinical evaluation of generative model based monitoring and comparison with compressive sensing

Research output: Chapter in Book/Report/Conference proceedingConference contribution

1 Scopus citations

Abstract

Generative model based resource efficient monitoring is an emerging data collection technique that has been shown to have compression ratio of around 40 in simulation environment on medical grade data from MIT BIH database. This paper discusses the intermediate outcomes of an ongoing clinical study where GeMREM enabled sensors are deployed on 125 subjects at the St Luke's cardiac hospital. According to the data from 25 patients we see that GeMREM achieves a compression ratio of 33, the reduction attributed to motion artifacts. We also compare the diagnostic accuracy of GeMREM with compressive sensing (CS) based ECG monitoring techniques. The results show that GeMREM although has better resource efficiency, CS is more accurate in representing temporal parameters such as heart rate, standard deviation of heart rate, and heart rate variability. However, interestingly, GeMREM is more accurate in preserving the shape of an ECG beat. Usage of dual basis in CS also cannot achieve shape accuracy comparable to GeMREM. Further, the reconstruction algorithm for GeMREM is almost 20 times faster than that for CS techniques.

Original languageEnglish (US)
Title of host publicationProceedings - Wireless Health 2015, WH 2015
PublisherAssociation for Computing Machinery, Inc
ISBN (Print)9781450338516
DOIs
StatePublished - Oct 14 2015
Event6th Wireless Health Conference, WH 2015 - Bethesda, United States
Duration: Oct 14 2015Oct 16 2015

Other

Other6th Wireless Health Conference, WH 2015
CountryUnited States
CityBethesda
Period10/14/1510/16/15

Keywords

  • Experimentation
  • Measurement
  • Performance
  • Reliability

ASJC Scopus subject areas

  • Computer Networks and Communications
  • Health Informatics
  • Health Information Management

Fingerprint Dive into the research topics of 'Clinical evaluation of generative model based monitoring and comparison with compressive sensing'. Together they form a unique fingerprint.

Cite this