
Electrocorticographic (ECoG) speech brain-computer interfaces (BCIs) show promise for restoring communication in amyotrophic lateral sclerosis (ALS), but it remains unclear whether chronic changes in signal quality are accompanied by loss of task-relevant neural information. We tracked signal characteristics and decoding over 25 months in a participant with ALS to determine how high-gamma (HG, 70–170 Hz) activity changes over time and whether these changes affect offline speech decoding. We implanted two 8 × 8 subdural ECoG grids over left sensorimotor cortex in a participant with slowly progressive bulbar variant ALS. Across 25 months, the participant performed an overt syllable-repetition task (12 consonant-vowel tokens) during simultaneous ECoG and audio recording. We quantified HG activation ratio (ActR), spectral signal-to-noise ratio (SNR), and peak z-scored HG responses. Speech acoustics were evaluated with triangular vowel space area (tVSA). Longitudinal trajectories and transition timing were estimated using generalized additive and piecewise models. A modified EEGNet-SE decoder was evaluated using 11 chronological rolling train-validation-test splits. We also assessed cross-temporal generalization, electrode saliency, neural class separability, and decoded mutual information. The neural metrics showed a common transition approximately 180 days after implantation. Before the transition, all three metrics increased. Thereafter, HG activation ratio remained stable (− 0.003 dB/100 days; P = 0.736), active SNR declined (− 0.473 dB/100 days; P < 10− 60), and z-scored HG peaks continued to increase slowly (0.067 per 100 days; P < 10− 20). tVSA showed mild nonlinear reductions. Across 726 held-out train–test evaluations, mean 12-syllable decoding accuracy was 74.6
Functional recovery after stroke is multifactorial, and predicting medium- to long-term outcomes remains challenging. This study applied machine learning (ML) to data from the prospective observational STRATEGY study to identify, among a minimum set of variables collected at admission to intensive inpatient rehabilitation after stroke, predictors of functional independence at discharge and at 3- and 6-month follow-up. We analysed data from adults admitted within 30 days of stroke to the Florence centre of STRATEGY. Disability in activities of daily living, measured by the modified Barthel Index (mBI) at discharge, at 3 and 6 months after stroke was the outcome of interest. Predictors comprised the minimal assessment protocol for stroke survivors including demographics, premorbid functional scores, neurological and functional variables collected at admission, and a set of markers of clinical complexity in rehabilitation such as dysphagia and urinary incontinence. Three ML algorithms were trained and validated within a nested cross-validation strategy. Performance was evaluated using Median Absolute Error (MedAE). Model interpretability used SHAP values to quantify each predictor’s contribution. Of the 244 recruited patients, 135 had complete outcome data at follow-up. XGBoost yielded the best performance (MedAE 7.83 at discharge, 9.15 at 3 months; 9.52 at 6 months). Baseline mBI was the strongest predictor across all timepoints. Besides mBI, the discharge model attributed greater predictive contributions to acute and subacute factors, including trunk control, neurological severity, lower limb strength, and age and time from the event, whereas 3-month prediction showed greater contributions from premorbid participation, lower limb strength, and cognitive status. At 6 months, neurological severity remained relevant, while premorbid mobility and disability, educational level, and comorbidity burden gained importance. ML models based on a minimal set of relevant clinical and anamnestic variables collected at admission provided accurate predictive performance at all time points. The results support the clinical value of easily and routinely collected variables at admission to inpatient rehabilitation, and show that prognostic factors are phase-dependent rather than static, with clinical and functional measures overarching at discharge, while premorbid disability, global cognitive function, and neurological severity gained importance at longer follow-up. NCT05389878.
Neuromodulatory interventions for dysphagia are increasingly explored as adjunctive rehabilitation strategies, yet their efficacy may be limited by non-individualized stimulation protocols and insufficient consideration of network-specific brain dynamics. Advances in neuroengineering now enable individualized targeting based on functional neuroimaging and computational modeling, offering the potential for precision neuromodulation of swallowing-related brain networks. We investigated whether individualized multichannel transcranial direct current stimulation (mc-tDCS), optimized using magnetoencephalography (MEG)-derived target localization, MRI-based finite-element head modeling, and a distributed constrained maximum intensity (D-CMI) optimization framework, modulates the cortical swallowing network. In a randomized sham-controlled crossover design, eight healthy adults underwent three stimulation conditions: low-intensity mc-tDCS (1.5 mA), high-intensity mc-tDCS (4.0 mA), and sham stimulation. Swallowing-related cortical activity was assessed before and after stimulation using MEG source reconstruction of event-related desynchronization during uncued voluntary swallowing and a visually cued swallowing task requiring precise timing. A repeated-measures Dose × Task × Time model tested the hypothesized interaction separately in five frequency bands, with Holm correction across bands. No Dose × Task × Time interaction was significant after Holm correction across the five frequency bands (all adjusted p ≥ 0.274). Within-condition pre–post analyses identified alpha- and beta-band changes after 1.5-mA stimulation during challenged swallowing and a low-gamma change after 4.0-mA stimulation during uncued swallowing. Challenged-task timing accuracy increased over time, without significant Dose or Dose × Time effects. This study demonstrates the feasibility of integrating individualized MEG-based targeting, subject-specific head modeling, and optimized multichannel stimulation. The hypothesized Dose × Task × Time interaction was not significant. Exploratory within-condition pre–post differences were observed in alpha, beta, and low-gamma activity, but do not establish stimulation-specific effects. Larger controlled studies are needed to evaluate the physiological effects of this individualized approach.
Exoskeletons are becoming increasingly popular solutions to promote motor recovery in patients with neurological disorders in rehabilitation. In the framework of human-device interaction, perceived user transparency plays a crucial role, as it allows users to move naturally and remain actively engaged during task execution. In this context, synergy-based approaches offer a valuable tool to gain a deeper understanding of the neural mechanisms underlying transparency and robot assistance. This study investigated the user transparency of the ALEx upper-limb exoskeleton and the effects of robotic assistance by means of the novel kinematic-muscular synergies, which describe how muscle synergies connect to the task space. Sixteen healthy subjects were recruited to perform a wide set of standardized arm movements in three modalities (Free, i.e., free movement, Transparent, i.e., robot provides no assistance, and Assisted, i.e., robot assists motion). Fifteen electromyographic (EMG) upper-limb and trunk muscles and trunk and upper-limb kinematics were collected to allow the extraction of kinematic-muscular synergies through the Mixed Matrix Factorization (MMF) algorithm. Averaged matched synergies were compared across modalities in terms of their spatial weights and temporal components to examine how the interaction with the exoskeleton altered the motor modules, providing a neurophysiological indicator of transparency. Six synergies were extracted from each modality to allow direct comparisons across conditions. The spatial and temporal structure of the synergies typical of Free movement were largely preserved when interacting with the device in the Transparent modality (similarities > 0.85 spatial, > 0.75 temporal), and in the Assisted (similarities > 0.90 spatial, > 0.70 temporal), even though a decrease in temporal magnitude (p = 0.004) was registered during the Assisted modality, indicating a reduction of the muscular effort. Our findings suggest that ALEx achieves a good level of user transparency and reproduces physiological motor patterns in the Assisted modality. The proposed MMF-based approach offers a promising tool for evaluating the impact of robotic devices on motor control, by explicating the link between the neural space, typically analysed in synergies, and the corresponding task performance.
Due to the increased stability demands of turning, the Figure-of-8 Test (F8T) may be a promising test to capture balance impairment after stroke. However, it remains unclear whether the F8T holds additional discriminative value after correcting for the influence of regular walking speed. This study evaluated the discriminative and construct validity of the F8T in people with stroke (PwS), accounting for the effects of walking speed. PwS with self-reported balance problems (n = 20) and healthy individuals of similar age (n = 20) performed the F8T, completing 4 laps around two hula hoops (d = 0.95 m) placed 3 m apart, both at comfortable and fast pace. Kinematics during the F8T were assessed using 3D motion capture. In addition, regular walking speed was obtained from a 2-minute walking test. Dynamic balance was evaluated using the miniBESTest. PwS showed a longer F8T time, larger center of mass trajectory length, longer turn duration, and less negative minimum mediolateral margin of stability compared to healthy individuals, both at comfortable and fast speed (all p’s < 0.001). These differences remained large and significant after correction for regular walking speed (p < 0.009). MiniBESTest scores were only associated with F8T performance at fast pace, showing a moderate association with F8T time (r=− 0.48, p = 0.031) and turn duration (r=− 0.49, p = 0.029), and strong association with the minimum mediolateral margin of stability during turning (r=− 0.60, p = 0.005). The F8T showed good discriminative abilities for balance impairment in PwS, even when corrected for the effects of a lower walking speed in PwS. The finding that PwS demonstrated less negative stability margins suggests that they adopted a more cautious movement strategy to avoid instability. Finally, the association between the F8T and the miniBESTest was only moderate to strong at fast pace, but should be interpreted in light of the relatively small sample size and limited range of balance problems within our stroke population.