Jove
Visualize
Contact Us
JoVE
x logofacebook logolinkedin logoyoutube logo
ABOUT JoVE
OverviewLeadershipBlogJoVE Help Center
AUTHORS
Publishing ProcessEditorial BoardScope & PoliciesPeer ReviewFAQSubmit
LIBRARIANS
TestimonialsSubscriptionsAccessResourcesLibrary Advisory BoardFAQ
RESEARCH
JoVE JournalMethods CollectionsJoVE Encyclopedia of ExperimentsArchive
EDUCATION
JoVE CoreJoVE BusinessJoVE Science EducationJoVE Lab ManualFaculty Resource CenterFaculty Site
Terms & Conditions of Use
Privacy Policy
Policies

Related Experiment Videos

Modeling electroencephalography waveforms with semi-supervised deep belief nets: fast classification and anomaly

D F Wulsin1, J R Gupta, R Mani

  • 1Department of Bioengineering, University of Pennsylvania, Philadelphia, PA, USA. wulsin@seas.upenn.edu

Journal of Neural Engineering
|April 29, 2011
PubMed
Summary

Related Concept Videos

You might also read

Related Articles

Articles linked to this work by shared authors, journal, and citation graph.

Sort by
Same author

Effectiveness of brainwave entrainment on pre-operative fear and anxiety in pediatric dental patients: a randomized controlled trial.

European archives of paediatric dentistry : official journal of the European Academy of Paediatric Dentistry·2024
Same author

Magnetic order and disorder environments in superantiferromagnetic [Formula: see text] nanoparticles.

Scientific reports·2022
Same author

The diabetes community exercise programme plus usual care versus usual care in patients with type 2 diabetes: A randomised, two-arm, parallel, open-label trial.

EClinicalMedicine·2022
Same author

Duration of EEG suppression does not predict recovery time or degree of cognitive impairment after general anaesthesia in human volunteers.

British journal of anaesthesia·2019
Same author

A biopsychosocial understanding of lower back pain: Content analysis of online information.

European journal of pain (London, England)·2017
Same author

A rare case of bullhorn-injury associated traumatic hernia of anterior abdominal wall managed by laparoscopic sutured tissue-only repair.

Journal of postgraduate medicine·2017

Deep belief nets (DBNs) offer faster, comparable classification of electroencephalography (EEG) data. Applying DBNs to raw EEG waveforms shows promise for automated recognition and anomaly detection.

Area of Science:

  • Neuroscience
  • Machine Learning
  • Signal Processing

Background:

  • Clinical electroencephalography (EEG) generates complex time-series data, primarily reviewed by human readers.
  • Deep belief nets (DBNs), a type of multi-layer neural network, are typically applied to image data, not time-series like EEG.
  • Automated analysis of physiological waveforms often relies on hand-crafted features, overlooking raw data potential.

Purpose of the Study:

  • To apply Deep Belief Nets (DBNs) in a semi-supervised learning paradigm for electroencephalography (EEG) waveform classification and anomaly detection.
  • To evaluate the effectiveness of DBNs using raw, unprocessed EEG data compared to traditional feature engineering.
  • To assess the speed and accuracy of DBNs for online automated EEG waveform recognition.

Main Methods:

Related Experiment Videos

  • Utilized a semi-supervised learning approach with Deep Belief Nets (DBNs) to model EEG waveforms.
  • Compared DBN performance against standard classifiers on an EEG dataset.
  • Investigated the use of raw EEG data versus hand-chosen features for classification and anomaly detection.
  • Leveraged the unsupervised learning component of DBNs to create an autoencoder for anomaly measurement.

Main Results:

  • DBN classification performance was comparable to standard classifiers on the EEG dataset.
  • DBN classification was significantly faster, ranging from 1.7 to 103.7 times quicker than other high-performing classifiers.
  • Using raw, unprocessed EEG data yielded comparable classification performance and superior anomaly measurement compared to hand-chosen features.
  • The autoencoder derived from DBN unsupervised learning effectively measured anomalies.

Conclusions:

  • Deep Belief Nets (DBNs) show potential for effective and efficient automated EEG waveform recognition.
  • Employing raw EEG data with DBNs can be a viable alternative to traditional feature extraction methods.
  • DBNs offer a promising approach for accelerating online automated analysis of complex clinical EEG data.