Search

During the past two decades several brain imaging studies have examined the neural correlates of disgust (the first study employing functional MRI was published by Phillips et al. (Phillips et al., 1997)). These studies usually combined the presentation of disgust-associated stimuli during neuroimaging acquisition and examined associated increases in regional brain activity. However, the original studies reported a number of different brain regions and it is still a matter of debate which regions may underlie disgust processing. On the one hand characteristics of the original studies such as low sample size, lack of power and robustness or differences in the recruited samples may have contributed to the inconsistent findings. On the other hand, differences in the employed stimuli may account for the inconsistent results. Among the previous studies, different kinds of stimuli have been employed, which could be classified into two main domains. The first domain contains facial expressions of disgust and film clips (also, static and dynamic disgust facial expressions) depicting disgusted facial expressions on the basis of their ability to evoke a vicarious experience of disgust, the second domain includes generally disgust-inducing pictures and film clips (also, static and dynamic disgust elicitors) which can evoke strong disgust-related feelings. The static and dynamic disgusted facial expressions carry strong socially communicative and interpersonal meaning and convey the experience of disgust of the depicted person to the observer whereas the disgusting stimuli induce directly a disgusting experience in the observer. It is therefore conceivable that the stimuli which vary along a social dimension (non-social, direct disgust inducing stimuli vs social communicative and interpersonal disgust stimuli) may induce differential as well as common neural activity. Despite a large number of disgust-related fMRI studies there is to our knowledge not a single neuroimaging meta-analysis that capitalized on the previous original studies to robustly determine the neural correlates of disgust processing or differences between the social and non-social disgust stimuli. To this end the present activation likelihood estimation (ALE) meta-analysis aims at investigating brain regions engaged during disgust processing in healthy subjects and to determine common and distinct neural correlates along the social and non-social dimensions. The ALE meta-analysis approach is not limited by experimental design, paradigm, task, number of subjects and data analysis methods (Eickhoff, Bzdok, Laird, Kurth, & Fox, 2012; Laird et al., 2005; Wager, Lindquist, & Kaplan, 2007), thus allowing general inferences to be drawn, which can help to resolve previous inconsistencies in the literature.

The specific neural systems underlying the subjective feeling of disgust remain debated. We therefore combine task-based functional MRI with multivariate pattern analysis (MVPA) to identify and evaluate a sensitive and generalizable neural signature to predict subjective disgust experience. The development of the decoder includes two stages, (1) acquisition of fMRI data during visual disgust processing in a discovery cohort – this data serves to train the neural signature, and (2) acquisition of fMRI data in a validation cohort using the same paradigm – this data serves as independent validation dataset to test the decoder. At the current stage we have finished the acquisition of the discovery cohort, the current preregistration serves to register the details of the acquisition of the validation cohort. During stage (1) from Oct.21-2021 to Mar.11-2022, we have finished the fMRI data acquisition of 80 healthy participants, which forms the discovery cohort. The subjects underwent an fMRI paradigm during which they were presented disgust inducing pictures. Following each picture subjects were asked to rate their current disgust experience on a 1-5 scale. Data were preprocessed according to our evaluated preprocessing protocols (Zhou et al., 2021, Nature Communications). Data from n = 2 subjects were excluded due to the excessive head motion during fMRI, thus leading to a final sample of 78 participants in the discovery cohort. The first-level fMRI analysis was next performed in the discovery cohort, which was used to obtain beta images for the prediction analysis. In this model, we included five separate boxcar regressors time-logged to the presentations of pictures for each rating (i.e., 1–5), which allowed us to model brain activity in response to each disgust level separately. For further details of the steps for conducting first-level fMRI analysis, please turn to one recent paper from our lab (Zhou et al., 2021, Nature Communications). Next, we applied whole-brain (restricted to a gray matter mask) MVPA to obtain a pattern of brain activity that best predicted participants’ self-reported disgust ratings. Specifically, we employed the support vector regression algorithm based on the individual beta maps (one per rating for each subject) as features to predict participants’ disgust experience. We employed a 10-fold cross-validation procedure (repeated 10 times) to evaluate the signature performance. Results showed that the developed visually induced disgust signature (VIDS) accurately predicted ratings of reported disgust in the discovery cohort. For instance, the average within-subject correlation between predicted and actual disgust ratings was r > 0.80, and the VIDS response accurately classified different levels of disgust experience (>80% accuracy). An important next step (stage 2) is to validate the VIDS performance in an independent dataset. To this end we are intending to collect n = 30 additional subjects with the identical fMRI paradigm. This data will serve as independent validation dataset. Following data acquisition, the data from this cohort will be processed using the identical procedures (e.g., preprocessing and subject-level general linear model analysis) as the discovery cohort. After obtaining the individual beta maps (one per rating for each subject), we will test the VIDS developed on the discovery cohort (with no further model fitting) in the validation cohort. We expect that, if the cross-validation procedure was not biased and the VIDS is robust and sensitive, comparable performances will be observed in the validation cohort (e.g., similar overall (within- and between-subject) and within-subject prediction-outcome correlations as well as comparable classification accuracies between different levels of disgust experience).