A novel and comprehensive artifact reduction strategy for EMG collected during fMRI at 3 Tesla

Functional MRI (fMRI) studies utilizing simultaneous and reliable electromyography (EMG) data can provide valuable functional output parameters to enhance image analysis and improve the study of brain function. Although strategies have been developed to reduce the noise in EMG signals collected during fMRI, few studies report rigorous validation or examine techniques in atypical EMG. In this work, a novel wavelet based artifact reduction tool is described. Stringent validation of the technique is performed in typical and fatigued EMG signals acquired from 10 typically developed subjects while collecting fMRI images at 3 Tesla. In the novel strategy, the signal is band-passed for the EMG spectrum. Periods of muscle activity are identified with a double threshold strategy based on amplitude of the signal. The stationary wavelet transform is employed using an 8 level analysis and thresholds optimized to remove artifact with minimal impact on EMG parameters. The artifact corrected EMG is analyzed traditionally for median frequency, power, and amplitude by root mean square. The technique removes visible artifact from the EMG in typical and fatigued conditions, amplitude was dampened with artifact correction and median frequency was not impacted by artifact correction. Post artifact correction, fatigued EMG was distinct from the unfatigued EMG, showing a significant decrease in median frequency ranging 3 to 10 Hz. These results demonstrate wavelet denoising as a viable artifact reduction tool for use with EMG collected during fMRI data acquisition. The work presents a method for tracking and identifying changes in muscle activity during image acquisition for fMRI studies.