Dual-task walking and automaticity after Stroke: Insights from a secondary analysis and imaging sub-study of a randomised controlled trial

Objective: To test the extent to which initial walking speed influences dual-task performance after walking intervention, hypothesising that slow walking speed affects automatic gait control, limiting executive resource availability. Design: A secondary analysis of a trial of dual-task (DT) and single-task (ST) walking interventions comparing those with good (walking speed ⩾0.8 m s−1, n = 21) and limited (walking speed <0.79 m s−1, n = 24) capacity at baseline. Setting: Community. Subjects: Adults six-months post stroke with walking impairment. Interventions: Twenty sessions of 30 minutes treadmill walking over 10 weeks with (DT) or without (ST) cognitive distraction. Good and limited groups were formed regardless of intervention received. Main measures: A two-minute walk with (DT) and without (ST) a cognitive distraction assessed walking. fNIRS measured prefrontal cortex activation during treadmill walking with (DT) and without (ST) Stroop and planning tasks and an fMRI sub-study used ankle-dorsiflexion to simulate walking. Results: ST walking improved in both groups (∆baseline: Good = 8.9 ± 13.4 m, limited = 5.3±8.9 m, Group × time = P < 0.151) but only the good walkers improved DT walking (∆baseline: Good = 10.4 ± 13.9 m, limited = 1.3 ± 7.7 m, Group × time = P < 0.025). fNIRS indicated increased ispilesional prefrontal cortex activation during DT walking following intervention (P = 0.021). fMRI revealed greater DT cost activation for limited walkers, and increased resting state connectivity of contralesional M1 with cortical areas associated with conscious gait control at baseline. After the intervention, resting state connectivity between ipsilesional M1 and bilateral superior parietal lobe, involved in integrating sensory and motor signals, increased in the good walkers compared with limited walkers. Conclusion: In individual who walk slowly it may be difficult to improve dual-task walking ability. Registration: ISRCTN50586966

2 relative changes in HHb and OHb concentrations. A low pass filter of 0.7Hz was used to remove high frequency noise and enable visual inspection of signals for motion artefacts. Blocks of data containing motion artefacts, missing signals or other noise were removed from analyses. Traces were then filtered with a moving Gaussian filter (1), using a width of 4s. Blocks included after motion artefact analysis and filter processing were detrended for the first 5 seconds preceding the task start and averaged for the task period and 20-second rest period after each task. For between subjects comparison, average signals for task and rest were normalized by dividing the whole average trace by the maximum concentration change within a channel and across tasks (2). The average relative concentration changes were calculated for the middle 10 seconds of both task and rest periods and used for statistical analyses.

Sup Figure 1 Anterior view of Prefrontal fNIRS optode placement.
Two detectors and light sources were used to create two channels which were placed over the left and right PFC covering an area between F7 and Fp1 and F8 and Fp2. Inter-optode-distance was 30mm.
Treadmill walking: Self-selected speed (fNIRS) To obtain selected walking speed the researcher exposed the stroke survivor to a range of speeds on the treadmill without giving feedback about the speed. The participants then chose the speed which they found comfortable and practised walking at this speed to confirm its selection and get familiarised to treadmill walking. If needed the person was allowed to hold one or both side-bars of 3 the treadmill whilst standing and walking. In some cases participants received some extra familiarisation sessions outside the assessment to get comfortable with treadmill walking. Processing: Brain extraction of the T1-weighted image was performed using OptiBET (Optimized brain extraction for patient brain) (3). The lesion was manually masked on the T1-weighted image using FSLview. Images were transformed to standard space (MNI 152), excluding the lesion region, using FMRIB's registration tools FLIRT (4) and FNIRT (5). individually went through the MELODIC output and identified components showing artefacts based on temporal and spatial features, following published protocols (6) . Noise components were removed from the fMRI data, then the de-noised datasets were entered into FEAT (7) for further analysis. Pre-processing included motion correction using MCFLIRT (8), non-brain removal using BET (9), spatial smoothing using a Gaussian kernel of 5 mm FWHM, grand-mean intensity normalisation and highpass filtering.

Task-fMRI
Statistical procedures were carried out in FEAT, with FMRIB's Improved Linear Model (FILM) (7). A boxcar regressor modelling the task and rest blocks was used to create first-level statistical maps, with each task modelled as a separate explanatory variable, for each participant at baseline and post-training separately. The "dual task cost" was determined by contrasting brain activation during dual task conditions with the sum of the activation during each of the corresponding single task conditions. For higher level analysis, EPI images were co-registered to the bias-corrected T1weighted image and then to MNI space using FLIRT.
Processing: EPI images were pre-processed using MELODIC, including motion correction using MCFLIRT (8), spatial smoothing using a 6 mm FWHM Gaussian kernel and temporal filtering. Noise components were classified manually (MF), according to published criteria (10) and removed using 5 FSL regfilt. The resulting de-noised images were co-registered to the bias-corrected T1-weighted image then transformed to MNI space, using FLIRT (8) and FNIRT (5) for further analysis.
Seed based connectivity analysis was conducted with primary motor cortex (M1) regions of interest (ROI), using the human motor area template (11). Dual regression was used to extract the timeseries from the ipsilesional and contralesional M1 ROI (in separate analyses) and correlate the timeseries with that of each voxel of the brain.

Pedal task (fMRI)
The pedal task required participants pedal (alternating dorsi-and plantar-flexion), each foot in opposite phase at a self-selected frequency on a purpose-built apparatus whilst lying on their back in the scanner (sup fig 2). An image of a moving foot was presented on the left side of the screen and

Picture planning (fNIRS and fMRI)
Picture planning task was a customised task, designed to test executive function in a manner that simulates distractions during community walking. Two pictures of the same daily activity were presented on the screen, and each picture represented a different action within the activity. The objective was to indicate whether the top or bottom picture came first in the sequence (sup fig 4).

Auditory Stroop (fNIRS)
The auditory variant of the Stroop task (12) required the participant to listen to the words "High" and "Low" spoken out at either a high pitch or a low pitch. The participant was asked to verbally indicate, by saying "high" or "low", whether the word was spoken out at a high or low pitch. The inter-stimulus-interval was set at 3.5s

The Number Stroop (fMRI)
The number Stroop variant of the Stroop task used visual stimuli presented on the screen; each stimulus consisted of a set of two numbers which were presented for 2 seconds. The numbers were presented vertically, one above the other, and differed in magnitude. In addition, the two numbers differed in font size. The objective was to indicate whether the top number or the bottom number was the biggest number in magnitude, irrespective of font size, using the rocker switch.

Intervention
Dual-task intervention schedule can be found in Supplementary table 2 8 Sup Sup Figure 5. fMRI Outcomes.