The impact of emotional valence and stimulus habituation on fMRI signal reliability during emotion generation

Alice Pirastru,Sonia Di Tella,Marta Cazzoli,Fabrizio Esposito,Giuseppe Baselli,Francesca Baglio,Valeria Blasi

DOI: https://doi.org/10.1016/j.neuroimage.2023.120457

IF: 5.7

2023-12-15

NeuroImage

Abstract:Background: The emotional domain is often impaired across many neurological diseases, for this reason it represents a relevant target of rehabilitation interventions. Functional changes in neural activity related to treatment can be assessed with functional MRI (fMRI) using emotion-generation tasks in longitudinal settings. Previous studies demonstrated that within-subject fMRI signal reliability can be affected by several factors such as repetition suppression, type of task and brain anatomy. However, the differential role of repetition suppression and emotional valence of the stimuli on the fMRI signal reliability and reproducibility during an emotion-generation task involving the vision of emotional pictures is yet to be determined. Methods: Sixty-two healthy subjects were enrolled and split into two groups: group A (21 subjects, test-retest reliability on same-day and with same-task-form), group B (30 subjects, test-retest reproducibility with 4-month-interval using two equivalent-parallel forms of the task). Test-retest reliability and reproducibility of fMRI responses and patterns were evaluated separately for positive and negative emotional valence conditions in both groups. The analyses were performed voxel-wise, using the general linear model (GLM), and via a region-of-interest (ROI)-based approach, by computing the intra-class correlation coefficient (ICC) on the obtained contrasts. Results: The voxel-wise GLM test yielded no significant differences for both conditions in reliability and reproducibility analyses. As to the ROI-based approach, across all areas with significant main effects of the stimuli, the reliability, as measured with ICC, was poor (<0.4) for the positive condition and ranged from poor to excellent (0.4-0.75) for the negative condition. The ICC-based reproducibility analysis, related to the comparison of two different parallel forms, yielded similar results. Discussion: The voxel-wise GLM analysis failed to capture the poor reliability of fMRI signal which was instead highlighted using the ROI-based ICC analysis. The latter showed higher signal reliability for negative valence stimuli with respect to positive ones. The implementation of two parallel forms allowed to exclude neural suppression as the predominant effect causing low signal reliability, which could be instead ascribed to the employment of different neural strategies to cope with emotional stimuli over time. This is an invaluable information for a better assessment of treatment and rehabilitation effects in longitudinal studies of emotional neural processing.