Table 1.
Participant demographics.
Fig 1.
Experimental protocol used to evaluate multi-session adaptation to multimodal biofeedback.
Participants completed a four-day protocol using combined audiovisual and sensorimotor biofeedback. Audiovisual biofeedback on soleus activity was provided unilaterally on the more-affected limb whereas sensorimotor biofeedback was administered bilaterally using a resistive ankle exoskeleton. Each session was separated into baseline (1 minute), biofeedback (2, 10-minute bouts), and washout (1 minute) phases. The nominal torque value of the ankle exoskeleton was set at 0.1 Nm/kg during the first bout of the first session and incrementally adjusted by 0.025 Nm/kg over the subsequent bouts, according to the schedule shown. Overground walking data were collected pre- and post-intervention. A licensed physical therapist also performed a full physical examination at the pre-intervention session. Motion capture data were collected during at the pre- and post-intervention sessions and electromyography (EMG) data were collected bilaterally from the vastus lateralis, semitendinosus, soleus, and tibialis anterior across all sessions.
Fig 2.
In-session soleus adaptation to multimodal biofeedback.
Median trend in peak soleus activity on the more-affected limb (left) and less-affected limb (right) for each minute of every ten-minute bout. Participants performed two 10-minute bouts in every session (1–4). The nominal resistance level applied by the sensorimotor biofeedback system was incrementally increased from 0.1 to 0.2 Nm/kg (normalized to participant bodyweight), according to the legend, where the left and right values under each session heading correspond to Bout 1 and 2, respectively. Soleus activity for each session was normalized to the 95th percentile of the in-session baseline walking phase. Both audiovisual (AV) and sensorimotor (SM) biofeedback were administered on the more-affected limb, whereas only SM biofeedback was administered on the less-affected limb so as to not impact user comprehension.
Fig 3.
Peak soleus activity across four sessions with multimodal biofeedback.
Peak soleus activity for the more-affected limb across all sessions with combined audiovisual and sensorimotor biofeedback. Each session was separated into two 10-minute bouts of biofeedback walking. Data represent activity from minutes 3–8 of each bout, defined as the mid-adaptation response. For each session, data were normalized to the 95th percentile of the one-minute baseline phase. A resistive ankle exoskeleton was used to provide sensorimotor biofeedback and the nominal resistance level was incrementally increased from 0.1 Nm/kg to 0.2 Nm/kg (normalized to participant bodyweight) across sessions, according to the figure legend. The left panel depicts group-wise trends, and the right panel shows individual responses as well as baseline soleus activity values (used for normalization).*indicates a significant change in soleus activity from baseline.
Fig 4.
Correlation between peak soleus activity during biofeedback and washout phases.
Mid-adaptation (minutes 3–8) peak soleus activity for each participant’s (n = 8) more-affected limb during the first (left) and second (right) bout of each biofeedback session compared to the washout phase. The washout phase represents the time at which biofeedback was turned off, which was used to evaluate short-term retention of in-session gains. All data have been normalized to the 95th percentile of baseline activity for each session. Individual dots represent participants and colors indicate the magnitude of the resistance applied bilaterally by the ankle exoskeleton (i.e., sensorimotor biofeedback), normalized to participant bodyweight.
Fig 5.
Correlation between peak soleus activity and peak vastus lateralis activity across biofeedback sessions.
Mid-adaptation (minutes 3–8) peak soleus activity for each participant’s more-affected limb compared with their peak vastus lateralis activity. Data has been normalized to the 95th percentile of baseline walking and separated into the first (left) and second (right) bouts for each session. Individual dots represent participants and colors indicate the magnitude of the resistance applied bilaterally by the ankle exoskeleton (i.e., sensorimotor biofeedback), normalized to participant bodyweight. Note that these trends represent six of the eight participants included in this study, as two (P6 and P7) had to be removed due to EMG signal loss.
Fig 6.
Functional measures of overground walking before and after multi-session adaptation to biofeedback.
Motor control complexity (tVAF1), gait deviation index (GDI), and walking speed (non-dimensional) were measured during overground walking before and after the biofeedback sessions. Bar plots represent the median (IQR) value across individuals and colored points represent participant data. No significant changes in walking function were observed following multi-session adaptation to combined audiovisual and sensorimotor biofeedback.