Table 1.
Patients’ demography.
Table 2.
Average monthly training.
Fig 1.
Brain-machine interfaces control strategy.
(A) We have developed and validated two BMI strategies: the hybrid EEG and EMG state machine control (HSM) and the EEG single leg control (SLC). Both strategies were used to control the actuation of a virtual avatar and a robotic gait device (Lokomat or exoskeleton). A portable haptic device was used to inform the user of the position of the virtual or robotic leg and the contact of these actuator’s feet with the floor in real time [35]. (B) For both BMI strategies, a 16 channel EEG cap was used. Electrodes were clustered over the arm area of the sensorimotor cortex for HSM and leg area for the SLC strategy. The ground and reference electrodes are reported in gray and light blue respectively. (C) The control strategy for the HSM (left panel) is based on navigation of a state machine (middle panel), using motor imagery, and a two-step EMG confirmation, using isometric muscle contraction of the biceps (IMC). For example, when the subject is in a standing position (Stop/Safe sate) and wants to start walking, (s)he will imagine moving the left arm and confirms the choice with a left bicep IMC. The subject has then to produce a right bicep ICM to trigger walking. The SLC strategy uses the decoding of leg motor imagery through EEG signals. If left motor imagery (LMI) is detected the left step is triggered. Once in this position, if right motor imagery is detected, the right step is triggered, and if no state is detected for 5 seconds, the actuator (the avatar or the robotic gait device) returns to the idle position.
Fig 2.
Sensory improvement following the WA-NR training protocol.
(A) Patients’ normal (dark pink) and altered nociceptive sensations (light pink) areas measured with the standard ASIA assessment [6] at the onset of the training. Dermatomes in dark blue represent body areas where patients recovered normal sensations after our training; areas in the light blue are those where patients gained altered sensation (reported after 12, 22 and 28 months of training). (B) Mean ± standard error of the mean (SEM) for gained dermatomes with normal (dark pink), altered (light pink) and zone of partial preservation (ZPP) with nociceptive sensation for GR1 patients except P2. (C) Mean±SEM of the nociception improvement rate during the period of the WA-NR (difference of score between the onset and the end of the training, normalized by the number of months between the two evaluations) compared to the mean improvement rate before the training (difference of score between baseline and onset measurement normalized by the number of months between the two measurements). (D) Means ± SEM for the gained sensory score for each assessment of GR1 patients for nociceptive sensation, crude touch, sensitivity to pressure and temperature compared to the onset of training. Patient P7 stopped the training after 12 months and was therefore reported separately from the group average (in red). The score obtained after 4 months of training is compared to the one registered after 12 and 28 months of training (t-test, * P<0.05, ** P<0.01, *** P<0.001).
Fig 3.
Proprioception and sensitivity to vibration.
(A) Mean±SEM for proprioception (B) for vibration score for GR1 (black) and patient P7 (red). (C) Eight stimulation areas for the vibration test. Patients were blindfolded during the exam and had to report if they could feel the vibration and report the location of the stimulation (rib, hip, knee, ankle or foot). (D) Vibration map for patient P6. A solid line connects the actual vibration position (left circle (R)ib, (H)ip, (K)nee, (A)nkle and (F)oot) and the felt vibration position (right circle) by the patient. A dashed line means that the patient indicated the contralateral leg. (E) Occurrences of confusion toward a proximal joint (e.g., stimulation was done in the knee and patient-reported he felt in the hip), distal joint or with a contralateral joint.
Fig 4.
Clinical evidence. (A) Five key lower-limb muscles (proximal and distal rectus femoris, tibialis anterior, extensor hallucis longus and gastrocnemius) were evaluated eight times throughout the training [6] (0 to 28 months) and before starting the training (B: baseline). The motor score describes the amplitude of a contraction from 0 (absence contraction) to 5 (normal contraction). (B) Motor score for seven non-key lower-limb muscles measured at the onset of the training (0) and the end of the end of the training (28 months) for all patients. (C) Number of patients (over eight) with present muscle contraction for three abdominal muscles and the anal sphincter muscle (myotome S4-S5). (D) The LEMS is obtained for each patient by summing the score of all key muscles reported in panel A bilaterally. Missing data periods are in black. (E) Mean± SEM of the motor score for GR1 patients is reported in black and motor score for patient P7 is in red.
Fig 5.
Neurological and neurophysiological evidence. (A) Example of active myotomes compared to the Anatomical Lesion (AL) shown for patient P4. For each muscle, the graphic shows the corresponding myotome level (considering the principal nerve root, S2 Table), and the clinical score at the onset (0) and the end (+28 months) of the WARN training. The MRI of this patient’s spinal cord revealed an AL extending between T7 and T9 segments. Accordingly, at the onset of the training, the patient had preserved motor functions in the upper-limbs and the upper abdomen muscle (spinal nerve roots are located at T7-T8 level), but could not contract the middle and lower abdomen muscles (T9-T12 segment) nor any of the lower limb muscles. After 28 months of the WANR training, this patient had recovered partial motor functions in the middle and lower abdomen as well as in multiple lower limb muscle innervated under the AL, namely, rectus femoris proximal (L2) and distal (L3), hip adductor (L2), gluteus medius/maximus, tibialis anterior (L4), and gastrocnemius (S1). B) For each patient, we considered the muscles where motor functions were clinically observed (ASIA motor score = >1) and calculated the distance to the AL. The distance was calculated as the number of myotomes between the spinal nerve root of the muscle and the lowest segment of the AL. A positive value in the graph, corresponds to a muscle that is rooted below the anatomical lesion; and negative values refer to muscles rooted above the lesion. We report results for the onset (0) and the end of the training (28). Trunk muscles are reported with an open circle, lower limb muscles with an open triangle and upper limb muscles are not considered. (C) EMG envelops for the gluteus maximums muscle for all patients. Patients were instructed to contract their legs for periods of 5 seconds over a 3-minute period. Verbal instructions were given to the patient by the PT; and instruction periods are shown in gray in the graph. A dark gray area highlights the trials where the patient had a significant GMx contraction (> mean + 3xSD of the baseline), and light gray indicates those where the contraction did not reach significance. Muscle responses are shown for all patients (P1 to P8) at an early stage and later in training.
Table 3.
ASIA sacral evaluations.
Table 4.
Intestinal function evaluation.
Table 5.
Genitourinary evaluations.
Fig 6.
Neurological improvement and correlation with training the protocol.
(A) The coefficient of correlation between sensory and motor neurological improvements and the number of training hours of active locomotion, BMI-based training, patients SCI height, time since lesion and patients ‘age. (B) AIS grade improvement at 28 months. Patient P7 stopped the training after 12 months. A follow-up measurement was done with this patient, 16 months after he stopped the training. (C) WHOQOL-BREF [36] score for four subdomains.
Table 6.
Mean ± SEM score for all physical domains of the WHOQOL-BREF [36] questionnaire.
Table 7.
McGill score.
Table 8.
Mean ± SEM score for all psychological domains of the WHOQOL-BREF [36] questionnaire.