Mindfulness meditation training alters cortical representations of interoceptive attention

Abstract

One component of mindfulness training (MT) is the development of interoceptive attention (IA) to visceral bodily sensations, facilitated through daily practices such as breath monitoring. Using functional magnetic resonance imaging (fMRI), we examined experience-dependent functional plasticity in accessing interoceptive representations by comparing graduates of a Mindfulness-Based Stress Reduction course to a waitlisted control group. IA to respiratory sensations was contrasted against two visual tasks, controlling for attentional requirements non-specific to IA such as maintaining sensation and suppressing distraction. In anatomically partitioned analyses of insula activity, MT predicted greater IA-related activity in anterior dysgranular insula regions, consistent with greater integration of interoceptive sensation with external context. MT also predicted decreased recruitment of the dorsomedial prefrontal cortex (DMPFC) during IA, and altered functional connectivity between the DMPFC and the posterior insula, putative primary interoceptive cortex. Furthermore, meditation practice compliance predicted greater posterior insula and reduced visual pathway recruitment during IA. These findings suggest that interoceptive training modulates task-specific cortical recruitment, analogous to training-related plasticity observed in the external senses. Further, DMPFC modulation of IA networks may be an important mechanism by which MT alters information processing in the brain, increasing the contribution of interoception to perceptual experience.

Fig. 1

Respiration signal analyses. (A) Mean respiratory frequency (Hz) as a function of attention condition, as derived from a breath counting algorithm. The MT group displayed slower respiration during IA relative to the untrained group, whereas the groups did not differ during EA. (B) Mean respiratory volume (arbitrary respiration belt units) as a function of attention condition, as derived from a breath counting algorithm. The MT group displayed greater respiratory volume (deeper breaths) during IA relative to the untrained group, whereas the groups did not differ during EA. (C) Rate/volume tradeoffs for respiration as a function of attention condition. The slopes did not differ as a function of attention condition or group.

Fig. 2

Insula attention activation by anatomical partition. A significant interaction between attention, anatomical partition and MBSR training was found, such that dorsal anterior insula IA activation was greater in the MT than untrained group. (A) A sagittal view of the eight anatomical ROIs drawn to fit each gyrus of the insula on a template brain. (B) Percent signal change plots for IA recruitment as a function of insula partition location. (C) Percent signal change plots for EA recruitment as a function of insula partition location. Error bars represent s.e. VAC, ventral accessory gyrus; VS, ventral short gyrus; PL, posterior long gyrus; AL, anterior long gyrus; PS, posterior short gyrus; MS, middle short gyrus; AS, anterior short gyrus; AC, accessory gyrus.

Fig. 3

DMPFC activity. The dorsomedial prefrontal cortex (DMPFC) was the only region wherein attention (IA vs EA) condition interacted with group (untrained vs MT). The bar graph displays DMPFC task activations relative to the within-participant baseline period.

Fig. 4

Primary interoceptive seed activity and connectivity. A posterior insula region whose activity was associated with variation in respiratory rate during IA but not EA was selected as a primary interoceptive seed region for further analysis. The 3 mm spherical interoceptive seed region proved to be sensitive to the IA vs EA contrast (A). Comparison of whole brain functional connectivity revealed greater interoceptive seed connectivity with the middle insula/putamen in the MT group relative to the untrained group, irrespective of attention condition (B). Group differences in the PPI analysis suggested that in the untrained group, the posterior and middle insula demonstrated greater seed connectivity during IA than EA, reaching levels of connectivity observed across conditions in the MT group (C). Scatterplot r-values display the Fisher Z transformed mean of the within-subject correlations for each group. The scatterplots themselves show TR by TR activation patterns from a single representative participant in each group, whose within-subject correlation best matched the mean within-subject correlation for the group.

Fig. 5

DMPFC seed connectivity. The DMPFC was used as a seed ROI in a psychophysiological interaction (PPI) analysis, and these PPI maps were compared between untrained and MT groups. IA relative to EA predicted negative insula connectivity, as a main effect in the right posterior insula, an effect that appeared to be driven by the MT group in both whole brain (A) and insula ROI analyses (B). Scatterplot r-values display the Fisher Z transformed mean of the within-subject correlations for each group. The scatterplots themselves show TR by TR activation patterns from a single representative participant in each group, whose within-subject correlation best matched the mean within-subject correlation for the group.

Fig. 6

Relationship between MBSR daily practice completion and attention-related brain activity. Within the MT group, percentage of practice completion for each participant was entered as a covariate into the IA vs EA design matrix. The brain maps display significant conjunction regions between attention-related activity and the practice completion related activity. (A) Areas of IA > EA activation that are related to practice completion. (B) Correlations between activity in the right insula anterior long gyrus anatomical ROI and practice completion. Variance accounted for by practice completion (R²) is displayed for attentional bias (IA–EA), as well as the simple effects of IA and EA separately.