Related Experiment Video
Updated: May 27, 2026
08:33
A Randomized, Sham-Controlled Trial of Cranial Electrical Stimulation for Fibromyalgia Pain and Physical Function, Using Brain Imaging Biomarkers
Published on: January 5, 2024
Detection of physiological noise in resting state fMRI using machine learning
Tom Ash1, John Suckling, Martin Walter
1Wolfson Brain Imaging Centre, University of Cambridge, Cambridge, United Kingdom. twja2@wbic.cam.ac.uk
Human Brain Mapping
|November 29, 2011
Summary
This study introduces a novel method using machine learning to predict cardiac and respiratory cycles from fMRI data. This technique effectively removes physiological noise, even without direct monitoring, improving fMRI data quality.
Area of Science:
- Neuroimaging
- Biomedical Engineering
- Data Science
Background:
- Functional magnetic resonance imaging (fMRI) is susceptible to physiological noise from cardiac and respiratory cycles.
- Accurate physiological data is crucial for effective noise reduction in fMRI.
- Existing methods for physiological noise removal often require complete and accurate monitoring data.
Purpose of the Study:
- To develop and validate a technique for predicting cardiac and respiratory phase from fMRI data.
- To assess the utility of predicted physiological phase in detrending fMRI data.
- To compare the performance of noise reduction using predicted versus recorded physiological data.
Main Methods:
- A multiclass support vector machine algorithm was employed for time-point-by-time-point prediction of physiological phase.
- The technique was evaluated in scenarios with subject-specific training data and group-level training data.
- Noise reduction efficacy was assessed using Fourier transforms and seed-based correlation analysis with the RETROICOR tool.
Main Results:
- Predictions showed a strong correlation with recorded physiological phase (median Pearson correlation up to 0.99 for cardiac, 0.83 for respiratory).
- Using predicted phase with RETROICOR yielded comparable noise reduction to using recorded phase.
- Similar noise reduction patterns in Fourier spectra and brain regions were observed for both methods.
Conclusions:
- The developed technique accurately predicts cardiac and respiratory phase from fMRI data.
- Predicted physiological phase is a viable alternative to recorded data for physiological noise removal using tools like RETROICOR.
- This method is particularly useful when direct physiological monitoring data is incomplete or unavailable.
