Functional neuroplasticity in chronic post-stroke aphasia following a singing intervention in a cross-over randomised trial

Abstract

Group-based singing has been shown to improve language outcomes and induce structural neuroplasticity in chronic post-stroke aphasia (PSA). However, the functional neuroplasticity changes induced by such interventions remain unknown. Here our main aim was to determine these changes using a cross-over randomised trial. Nineteen patients with PSA were randomly allocated to a 4-month multicomponent singing-based intervention (singing group) or standard care (control group). With a pre-post design, we pooled data from both groups and analysed verbal learning and task-based fMRI activation of two novel songs (trained or untrained during intervention) at two time points (pre- and post-intervention). Post-intervention, patients with PSA produced more correct syllables from the trained song and for the trained relative to the untrained song. fMRI results revealed increased activation when singing along to the trained song in the right postcentral gyrus, and in the right posterior superior temporal gyrus (pSTG) when singing along to the trained vs. untrained song. Notably, right pSTG activation increases correlated with improved naming abilities. Collectively, these findings indicate that group-based singing is associated with verbal learning and induces functional neuroplasticity changes in the singing network, derived from demographically matched healthy controls, which are associated to improved naming abilities in chronic PSA. https://www.clinicaltrials.gov , Unique identifier: NCT03501797.

Fig. 1

Flowchart.

Fig. 2

Lesion overlap, experimental design and verbal learning results. (a) Patients with post-stroke aphasia (PSA) were divided into two intervention groups: A and B, according to the cross-over design of this RCT. Lesion overlap for the patients with complete fMRI acquisition at all time points (N = 19) rendered on the MNI template (see Methods for more details). (b) Sparse sampling trial design in the fMRI singing task. Each trial started with 0.5s of silence, followed by a period of inhalation (1.5s) and the active condition (4s), a new silence period (1.5s) and volume acquisition (1.5s). The design involved 3 active conditions for each phrase: listen, sing along and sing from memory. Additionally, 2 rest trials (exhale, dashed line) were included at the end of every three phrases. In total, the task comprised 110 trials distributed in 5 blocks that included 18 trials of an active condition and 4 rest trials. All patients underwent 3 fMRI sessions with a singing task: baseline (T1), 5-month (T2) and 9-month (T3). (c) Verbal learning was measured by the number of correct syllables. Pre- vs. post-intervention effect for the trained song (top row) and for the trained vs. untrained song (bottom row) after pooling data from both Groups and time points. (**): p < 0.01.

Fig. 3

Functional neuroplasticity changes induced by the singing intervention. (a) Singing network activation in the demographically matched healthy controls group (N = 30) for the sing along vs. rest contrast. This activation map was thresholded at FWEc at the cluster level and used as an explicit mask in the second-level analysis for the fMRI data in patients with PSA. (b) For the sing along vs. rest contrast, the right PoCG was more strongly activated for the trained song in the post- vs. pre-intervention comparison pooling data together from both Groups. (c) When comparing the trained vs. the untrained song, the right pSTG showed increased activation. Colour bars indicate the T value thresholded at p_unc = 0.005 for visualisation purposes. Abbreviations: PoCG, postcentral gyrus; pSTG, posterior superior temporal gyrus.

Fig. 4

Spearman’s correlation between intervention-induced functional neuroplasticity and improved naming abilities. The mean contrast estimate for the right pSTG was positively correlated with WAB Naming Change score post- vs. pre-intervention (p = 0.038) after partialling out the effect of age, total intracranial volume and lesion size. Translucent bands around the fitted regression line denote the 95%CI. Abbreviations: pSTG, posterior superior temporal gyrus; R: Spearman’s correlation coefficient.