AI-Guided Multi-Electrode Transcutaneous Spinal Cord Stimulation for Gait Rehabilitation in People With Spinal Cord Injury or Stroke

tSCS is a non-invasive technique that delivers electrical pulses through electrodes placed on the skin over the spine to modulate spinal neural circuits, facilitate the recruitment of lower-limb muscles, and support the recovery of voluntary movements. Conventional tSCS typically uses a single stimulation electrode, which may limit selective recruitment of different muscle groups. This study investigates a personalized multi-electrode approach to selectively recruit specific lower-limb muscle groups and deliver stimulation at appropriate phases of the gait cycle, with the aim of improving walking function.

The main questions this study aims to answer are:

1. How do electrode positions and stimulation intensities affect selective muscle recruitment, and how do responses vary between participants and days? 2. Can ML-based optimization identify stimulation settings with selectivity comparable to complete mapping of muscle responses, using fewer tests and less time? 3. Does personalized multi-electrode stimulation, synchronized with specific gait phases, improve walking function more than single-electrode tonic stimulation when both are combined with gait training?

The study includes 24 participants (14 with spinal cord injury and 10 with stroke) and consists of two stages.

Stage 1 - Characterization of tSCS Selectivity and Stimulation Optimization

All participants complete two experimental sessions while lying down. Researchers compare muscle responses to stimulation at a central electrode and at multiple spinal electrode locations. Electrical responses in the leg muscles are recorded using surface electromyography (EMG), and spinal reflexes are assessed using paired electrical pulses.

By systematically testing electrode positions and stimulation intensities, researchers build a map of which leg muscles respond to each setting (stimulation landscape). They assess the selective recruitment of different muscle groups, compare stimulation locations, and examine differences between participants and across the two sessions. They also test whether ML-based optimization can find similarly selective settings more efficiently than the complete mapping procedure.

Stage 2 - tSCS Combined With Gait Training

Participants are randomly assigned to one of two groups: single-electrode or multi-electrode tSCS, both combined with gait training. During spinal cord injury recruitment, the randomization procedure was adapted to better balance participants with different baseline walking abilities across groups.

Before each training session, participants lie down for an individualized calibration. An ML-based optimization method uses recorded muscle responses to select electrode positions and stimulation intensities for the subsequent walking session.

During walking, the single-electrode group receives continuous stimulation through one selected electrode. In the multi-electrode group, stimulation is delivered through electrodes selected to preferentially recruit different target muscle groups. A predefined walking-cycle model determines when stimulation should be delivered, based on gait phases detected in real time by wearable movement sensors.

Both groups complete 20 supervised training sessions over four weeks, including 30 minutes of treadmill walking per session. Walking speed is selected according to each participant's comfort and ability, with therapist adjustments, body-weight support, and physical assistance as needed.

Clinical and functional assessments are conducted before the intervention (T0), after the 20 sessions (T1), and at a 30-day follow-up (T2). They examine walking performance, muscle strength, spasticity, balance, and mobility. Additional measurements assess muscle activation, brain activity, and biological markers associated with neuroplasticity. The follow-up assesses whether changes are maintained after training.