Aberrant modulation of brain activity underlies impaired working memory following traumatic brain injury

Fig. 1. Flowchart of participants involved with each component of the study. All participants completed in the behavioural working memory tasks (N = 81); however, only a subset of participants underwent fMRI scanning whilst completing this task (i.e. 18 chronic TBI and 17 healthy controls).

Fig. 2. Schematic diagram of low and high cognitive load trials of the Sternberg working memory task. Two (low cognitive load) or six (high cognitive load) letters were presented during the encoding phase, followed by a maintenance phase in which the screen was blank, and finally a probe was presented during retrieval phase. Low and high cognitive load conditions were presented alternately.

Fig. 3. Behavioural results for the Sternberg working memory task. (A) Plots depicting accuracy, as measured using dprime (higher values denote better performance). Individual datapoints are displayed along with violin plots and boxplots showing distribution. Performances were comparable across groups in the low cognitive load condition (P > 0.05; panel A, left). In the high cognitive load condition, however, performance was significantly impaired for both the acute (P = 0.006) and chronic TBI groups (P = 0.024; panel A, right). (B) Plots representing reaction time (higher values denote slower performance). The acute TBI group was significantly slower than healthy controls in both the low (P < 0.001; panel B, left) and high cognitive load condition (P = 0.001; panel B, right). There were no significant differences in reaction time between the chronic TBI group and healthy controls in both load conditions (P > 0.05; panel B, left and right). Note: P-values were adjusted for multiple comparisons.

Fig. 4. Activation during the Sternberg working memory task for the fMRI cohort (18 chronic TBI, 17 healthy controls). (A) Pattern of activation segregated according to encoding, maintenance, and retrieval stages (red/yellow). There was minimal overlap between these clusters and TBI lesions (green). All subcomponents were associated with increased activation in frontal and occipital areas, the insula, and cerebellum. Encoding was associated with increased activation in temporal areas (e.g. left middle temporal gyrus, right superior temporal gyrus). Retrieval was associated with increased activation in the left angular gyrus. (B) Overall activation during the Sternberg working memory task (red/yellow) overlaid on a working memory meta-analytic mask (blue) derived from neurosynth. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

Fig. 5. fMRI group differences on the Sternberg working memory task. There was minimal overlap between significant clusters (red/yellow) and TBI lesions (green). (A) Average across all trials (i.e. regardless load and subcomponent), the TBI group showed reduced activation in the left cerebellum, left middle frontal gyrus, and left superior temporal gyrus. Panel A right, boxplot overlaid with individual datapoints comparing signal change when averaged across all trials between the TBI group and healthy controls. (B) The TBI group showed reduced activation in the low cognitive load condition in the left cerebellum, right middle frontal gyrus, left superior temporal gyrus, and left superior frontal gyrus. Panel B right, comparison of signal change during the low cognitive load condition between the TBI group and healthy controls. (C) With increased cognitive load, the TBI group showed attenuated reduction in activity in the right superior parietal gyrus, right inferior temporal gyrus, right middle frontal gyrus, and left inferior occipital gyrus. Panel C middle, comparison of signal change in the low and high cognitive load conditions between the TBI group and healthy controls. Panel C…