Introduction: Machine learning (ML), a subset of artificial intelligence, represents a promising method for decision-making under uncertain conditions. It can detect patterns in complex data, determine relationships between variables, and no requirement for prior knowledge as superior to conventional regression techniques. Due to the clinical complexity of spinal cord injury (SCI) patients, ideal motor rehabilitation therapy is hard to determine. The transcranial direct current stimulation (tDCS) and robotic gait training have shown potential benefits, but the high variation of response reduces their clinical translation; thus, a predictive model of response could help us to identify good responder candidates and to guide the resources allocation better. Therefore, we tested the hypothesis that a ML algorithm may predict the effectiveness of non-invasive brain stimulation combined with robotic gait training for motor improvement in SCI patients prior to treatment.
Methods: We applied a supervised ML approach to develop a classification model to predict motor function response in SCI patients after tDCS combined with robotic gait training. Patients undergone the treatment as part of the previous randomized clinical trial (NCT02562001). The primary outcome was motor function improvement (indexed by WISCI-II scale). Demographical (age and sex), clinical (AIS scale, level of lesion, pain status, medication history, rehabilitation treatment history, and time since injury), and electroencephalography variables (absolute and relative delta, theta, alpha, and beta band power from frontal, central and parietal areas) were included as predictors. An ML approach of feature engineering, data pre-processing, and model optimization was used to create the most accurate predictive model. A principal component analysis (PCA) algorithm was used to obtain the most differentiable features, which are used as features to train a Multilayer Perceptron (MLP) classifier based on a backpropagation learning algorithm. We used MATLAB ML toolbox to perform the modeling.
Results: This analysis included 39 incomplete SCI patients (AIS C or D in ASIA classification – 54% patients AIS D and 46% patients AIS C), 64% were paraplegic, 23 % female, mean age of 37.1 (SD ±13.5) years old, with average time since the injury of 15.4 (SD ±10.0) months, 56.4% presented chronic pain. We included 19 and 20 patients from active and sham group, respectively – four patients from the original clinical trial were excluded due to non-availability of the EEG recordings. Groups were relatively homogeneous for age, and time since injury, AIS scale, and WISCI-II classification at baseline. We included 30 and 9 subjects to train and test the model, respectively. The best performing predictive model, used an MLP classifier and had an average area under the curve (AUC) of 0.867, classification accuracy of 80%, sensitivity of 78.3%, and specificity of 81.3%. The relative beta band power over the central and parietal areas, the AIS scale, and pain status were associated with motor function improvement.
Conclusions: We developed a model that predicted positive motor outcome after tDCS combined with robotic gait training with good accuracy. Our analysis suggests the applicability of machine-learning as predictive modeling method for individualized treatment approach in SCI rehabilitation.

INTRODUCTION:
Despite an enormous amount of research, the prognosis for recovery of motor functions in patients with chronic motor completeSpinal Cord Injury (SCI) remains dismal.Spinal Epidural Electric Stimulation (sEES) has been the latest hope for SCI individuals with a few clinical studies, including ours recently showing positive results with regards to lower limb motor function after sEES.However, despite the consequences of cervical spine injuries being more devastating, only one case series has been published till date on cervical sEES.
As an extension to our previous trial, we aimed to assess the safety and efficacyof cervical sEES after SCI .
METHODS:
After Ethical Committee approval, we conducted a randomised pilot clinical study in six chronic motor-complete spinal cord injured patients who fulfilled the eligibility criteria. All the subjects provided a written as well as video informed consent. They were randomized into two groups of three patients each using simple randomisation (random number sequence). Each group underwent institutional standardised activity-based training. One group was implanted with asEES in addition (RESTORE ADVANCED,Medtronics).
The primary outcome measure included assessment of functional capabilities of hand and arms using CUE (Capabilities of Upper Extremity test) and GRASSP scale and assessment of motor power of upper limbs through ASIA grading (Upper Extremity Motor Score) with and without stimulator ON. The secondary outcome measures included assessment of any voluntary movement by additional muscle groups not included in ISNCSCI, Spinal Cord Indepence Measure (SCIM), assessment of respiratory function using Pulmonary Function Tests/weaning off ventilator, modified ashworth scale (MAS), WHO Quality of Life (WHOQOL-BREF) questionnaire and evaluation of bladder function by Urodynamic studies. All these scales are assessed every 6 weekly except urodynamic studies which were performed 3 monthly. Any adverse effects related to the procedure were noted and addressed.
Evaluation was also done to assess if the motor improvements were sustained when the stimulator was turned off.
RESULTS:
All the six subjects were either AIS-A(n= 4) or B(n=2) and were recruited atleast18 months post injury. The neurological level was C5 or above.Interim analysis at 12 weeks revealed that participantsin the sEES group have improved CUE, GRASSP scores,ASIA motor scores(with stimulator on), SCIM score, WHOQOL and improved respiratory function(SBC) compared to those with activity based training alone. ASIA grading (with stimulator off), MAS and urodynamic studieshave not shown significant difference between the groups at present.The participants are under long term follow up. The six month follow up results of all participants will be presented.
CONCLUSIONS:
To our knowledge, this is the first randomised clinical trial to assess the role of Cervical sEES. Cervical spinal epidural stimulation is a safe and efficient technique and has the potential to improve the motor and respiratory functions of patients with chronic SCI.However long term studies with more patients need to be done to draw a definite conclusion in this regard.

Introduction:
Based on suggestions for improving the Spinal Cord Independence Measure third version (SCIM III), we developed a fourth version of SCIM (SCIM IV). The most significant changes are the collapsing of the "transfer" items into one item and the discrimination between walkers and wheelchair users. We present here an international study of the validity and reliability of SCIM IV, and compare it with SCIM III.
Methods:
Data of 648 patients with spinal cord injury (SCI) were collected in 19 SCI Units from 11 countries. All patients were evaluated with SCIM IV and SCIM III, mostly by observation, by two raters, within a week after admission to the unit and before discharge. Conventional and Rasch analyses were used to assess the results.
Results:
The patients were 51±17 years old, 72% males, 43% with tetraplegia, 31% with AIS grade A, and 57% with traumatic SCI. Reliability: Total agreement between raters at admission, on the various SCIM IV scores for each subscale was 74%-94%, and kappa values were 0.634-0.790. Pearson correlation between raters’ scores was high for all subscales (r=0.898-0.982), and paired t-test did not reveal significant differences in the total SCIM IV score on most subscales (p>0.05). ICC at admission for SCIM IV subscales and total scores was ≥0.898. Cronbach's alpha values were 0.913, 0.658, and 0.917 for Self-care, Respiration and sphincter management, and Mobility, and it increased if Feeding, Respiration, Mobility in bed, or Stair management, but not other items, were deleted (α= 0.918, 0.723, 0.923, and 0.926, respectively). Rasch analysis showed a reliability index of 0.761-0.936 for the various subscales, and identified 2-5 discernible ability levels. No significant differences were found in item difficulty hierarchy between contrasted groups of age, gender, etiology, SCI level, or country. Validity: SCIM IV and SCIM III scores were very close and highly correlated (r=0.861-0.963, p<0.001). They differed slightly, and the difference was not significant for the total SCIM scores. In most comparisons, the responsiveness of SCIM IV was not found to be different from that of SCIM III (p>0.05). “Rasch ruler” showed that on all SCIM IV subscales the distribution of item difficulty grades matched the distribution of person ability measurements. Most SCIM IV items showed satisfactory fit to the Rasch model (MNSQ fit index 0.7-1.4). SCIM subscales were found to be unidimensional: the percentage of variance explained by the Rasch measure of person ability and item difficulty of the total variance found in the data was 99.5% for Self-care, 98.5% for Respiration & sphincter management, and 99.8% for Mobility. SCIM IV item characteristic curves showed that 85% of the categories of SCIM IV items were properly ordered (emergent).
Conclusions:
SCIM IV is reliable and valid, and it is suitable for assessment of daily activities of SCI patients. The mean value of the scores of all items, as a percent of their maximal scores, was very close to those of SCIM III, which makes possible appropriate comparisons with SCIM III in future group studies.

Introduction:
Activity-based therapy (ABT) is defined as therapeutic activities that involve “repetitive neuromuscular activation below the level of spinal injury, typically achieved through intensive, task-specific movement practice.”1 Technologies, such as functional electrical stimulation and virtual reality, have been developed to assist the delivery of ABT in clinical environments. However, the implementation of ABT and its associated technologies in clinical practice have encountered challenges and little is known about the barriers or facilitators experienced by clinicians who aim to provide ABT to individuals living with spinal cord injury (SCI). The purpose of this study is to understand if and how ABT and its associated technologies are being used in community-based settings in Canada.

Methods:
In this qualitative exploratory study, we targeted nine community-based clinics across Canada that provided ABT programs to individuals living with SCI. Occupational therapists, physical therapists, kinesiologists and clinic managers or owners were interviewed by the researchers to describe their experiences with ABT and the technologies associated with it for individuals living with SCI. Semi-structured questions queried what technologies or techniques were being used to assist ABT and why, and what barriers and facilitators to providing ABT exist. Interpretive description was used to analyze the transcribed interviews and to derive themes.

Results: Three physical therapists, one occupational therapist, seven kinesiologists and one clinic owner from nine clinics in four different provinces in Canada completed the interview (i.e. 12 participants in total). Six major themes were identified with the analysis: 1) There is variety in therapy (i.e. ABT can be done with and without technology, a large variety of equipment can be used, there is a large variety of approaches), 2) ABT is focused on the clients (i.e. value of emotional well-being and personalized goals), 3) Professional background and continuing education influence the ABT program provided, 4) Interventions should maximize neuroplasticity (i.e. combination of types of sensory feedback, many repetitions, weight-bearing, technology enhances therapy), 5) There are many barriers to provide ABT in the community (i.e. cost of technology and therapy, need to maintain operations of a private practice , need for space) and 6) The landscape of ABT in Canada is positively changing (i.e. relationship with public health care services is improving, research on ABT is increasing, need for education for therapists and for individuals living with SCI).

Conclusions: The use of ABT and associated technologies varied across community-based clinics in Canada based on the background and training of the therapists. However, the client’s specifics goal and well-being were consistently the focus of the ABT programs, and the importance of maximizing neuroplasticity was consistently acknowledged. Many barriers to the use of ABT and its associated technologies were identified, but participants were optimistic about the future use of ABT for Canadians living with SCI.

1. Musselman KE, Shah M, Zariffa J. Rehabilitation technologies and interventions for individuals with spinal cord injury: translational potential of current trends. Journal of NeuroEngineering and Rehabilitation 2018; 15: 40.