Development of a methodology to combine fMRI and EMG to measure emotional responses in patients with anorexia nervosa

Abstract

Objective: Individuals with eating disorders are theorized to have basic impairments in affective appraisal and social-emotional processing that contribute to pathogenesis of the disease. We aimed to determine if facial electromyography could be used to discriminate between happy and disgust emotions during simultaneous acquisition of an fMRI BOLD sequence in efforts to establish a novel tool for investigating emotion-driven hypotheses about eating pathology. In line with standards for rigor and reproducibility, we provide detailed protocols and code to support each step of this project.

Method: Sixteen adolescents with low-weight eating disorders viewed emotional faces (Happy or Disgust) and were asked to mimic their facial expression during simultaneous BOLD and EMG (Corrugator supercilli, Lavator lavii, Zygomaticus major) acquisition. Trials were repeated with the scanner off and again with scanner on (i.e., fatigue).

Results: The Levator and Zygomaticus activation patterns discriminated disgust and happy faces successfully. The pattern held between scanner on and off conditions, but muscle activation attenuated in the Fatigue condition, especially for the Zygomaticus.

Discussion: Simultaneous fMRI-EMG is a new tool capable of discriminating specific emotions based on muscle activation patterns and can be leveraged to answer emotion-driven hypotheses about clinical populations characterized by difficulty labeling or processing emotions.

Trial registration: ClinicalTrials.gov NCT02795455.

Keywords: adolescent; anorexia nervosa; disgust; electromyography; emotional processing; fMRI.

Conflict of interest statement

Conflicts of Interest: The authors have no conflicts of interest to declare.

© 2018 Wiley Periodicals, Inc.

Figures

Figure 1a-c.. Raw and Filtered Singal of Levator Lavii Muscle for Single Subject.

The blue timseeries signal was sampled at 10,000 Hz and displays of the full 10 trial acquisition block for the scanner condition ‘On’. The Red time series signal included a low-pass filter at 60 Hz and notch filter at 10 Hz to account for the TR signal. The Yellow time series used a band pass filter of 20—60 Hz. Black lines represent the TR signal and the Green markers are the Eprime markers for 4 second epochs of stimulus presentation alternation “Smile” then “Yuck”.

Figure 1a-c.. Raw and Filtered Singal of Levator Lavii Muscle for Single Subject.

The blue timseeries signal was sampled at 10,000 Hz and displays of the full 10 trial acquisition block for the scanner condition ‘On’. The Red time series signal included a low-pass filter at 60 Hz and notch filter at 10 Hz to account for the TR signal. The Yellow time series used a band pass filter of 20—60 Hz. Black lines represent the TR signal and the Green markers are the Eprime markers for 4 second epochs of stimulus presentation alternation “Smile” then “Yuck”.

Figure 1a-c.. Raw and Filtered Singal of Levator Lavii Muscle for Single Subject.

The blue timseeries signal was sampled at 10,000 Hz and displays of the full 10 trial acquisition block for the scanner condition ‘On’. The Red time series signal included a low-pass filter at 60 Hz and notch filter at 10 Hz to account for the TR signal. The Yellow time series used a band pass filter of 20—60 Hz. Black lines represent the TR signal and the Green markers are the Eprime markers for 4 second epochs of stimulus presentation alternation “Smile” then “Yuck”.

Figure 2.

The interaction demonstrates that Zygomaticus and Levator clearly descriminate between faces in the hypothesized directions. Zygomaticus was significantly greater in the Smile condition and Levator was significantly greater in the Yuck condition. Error bars are standard error of model estimated means.

Figure 3.

The interaction demonstrates that the condition did not yield systematic differences in Levator and Corrugator signals but that there is a systematic increase in Zygomaticus activation yielded significant differences between Fatigue and scanner On activation blocks. Although the other patterns appear variable, there were no significant differences between muscles across conditions. Error bars are the standard error of the model estimated means.

Figure 4.

Blood oxygen level-dependent (BOLD) signal increases for smile and yuck face simulations in adolescent females with low-weight eating disorders. Smile and yuck faces produced similar patterns of BOLD signal changes in anterior cingulate, inferior frontal, anterior insula, and parietal regions, and differential signal changes in ventromedial prefrontal cortex and amygdala. Figures are thresholded at p < 0.01 (corrected for multiple comparisons with a cluster threshold > 25 voxels). Numbers at bottom indicate z coordinates in the Montreal Neurological Institute brain template space. L = left, R = right.