Prediction of robotic neurorehabilitation functional ambulatory outcome in patients with neurological disorders

 Why would any survivor think that predictions of failure to recovery have any meaning at all?   Does no one in stroke ever think about what the survivor experience is like? Anything less than 100% recovery is failure. Failure should be rewarded by firings.

Prediction of robotic neurorehabilitation functional ambulatory outcome in patients with neurological disorders

• Chao-Yang Kuo,
• Chia-Wei Liu,
• Chien-Hung Lai,
• Jiunn-Horng Kang,
• Sung-Hui Tseng &
• Emily Chia-Yu Su 

Journal of NeuroEngineering and Rehabilitation volume 18, Article number: 174 (2021) Cite this article

• 133 Accesses

• 2 Altmetric

• Metrics details

Abstract

Introduction

Conflicting results persist regarding the effectiveness of robotic-assisted gait training (RAGT) for functional gait recovery in post-stroke survivors. We used several machine learning algorithms to construct prediction models for the functional outcomes of robotic neurorehabilitation in adult patients.

Methods and materials

Data of 139 patients who underwent Lokomat training at Taipei Medical University Hospital were retrospectively collected. After screening for data completeness, records of 91 adult patients with acute or chronic neurological disorders were included in this study. Patient characteristics and quantitative data from Lokomat were incorporated as features to construct prediction models to explore early responses and factors associated with patient recovery.

Results

Experimental results using the random forest algorithm achieved the best area under the receiver operating characteristic curve of 0.9813 with data extracted from all sessions. Body weight (BW) support played a key role in influencing the progress of functional ambulation categories. The analysis identified negative correlations of BW support, guidance force, and days required to complete 12 Lokomat sessions with the occurrence of progress, while a positive correlation was observed with regard to speed.

Conclusions

We developed a predictive model for ambulatory outcomes based on patient characteristics and quantitative data on impairment reduction with early-stage robotic neurorehabilitation. RAGT is a customized approach for patients with different conditions to regain walking ability. To obtain a more-precise and clearer predictive model, collecting more RAGT training parameters and analyzing them for each individual disorder is a possible approach to help clinicians achieve a better understanding of the most efficient RAGT parameters for different patients.

Trial registration: Retrospectively registered.

Introduction

Neurological disorders are often chronic and debilitating, and place heavy burdens on families and society [1]. Improving mobility is one of the main goals of rehabilitation for patients with neurological disorders [2]. In neurorehabilitation, a high dose and intensity, sufficient practice, individualized goals, motivation, and specialist knowledge are all important factors for achieving better outcomes [3, 4]. Compared to conventional therapy, robotic-assisted gait training (RAGT) is expected to more effectively improve mobility, as it can provide a higher dose and more-intensive treatment than usual rehabilitation [5]. As early as 2009, randomized controlled trials showed that RAGT combined with regular physiotherapy was more effective for improving the functional ambulation capacity and neurological recovery than conventional therapy in patients after a subacute stroke [6].

Comparisons of the efficacy between RAGT and conventional gait training (CGT) have garnered considerable attention in rehabilitation medicine. A review article revealed that RAGT applications demonstrated a better effect than CGT in post-stroke patients [7]. Another review article provided evidence that RAGT improved walking function in patients that sustained a spinal cord injury within the past 6 months [8]. In contrast, other studies demonstrated non-superior results of the effectiveness of RAGT for functional recovery of walking in survivors with different neurological disorders [9, 10]. Variations in the intensity, duration, and amount of training, as well as in the types of treatment, participant characteristics, and measurements across trials may have contributed to different reported effectiveness levels. Nevertheless, specific applications using RAGT devices to obtain optimal effects for patients remain unclear.

Robot-assisted gait trainers operate by either end-effector (non-portable) or exoskeleton (portable) principles. One systematic review showed that operational robots, such as the Gait Trainer, were more cost-effective in achieving independence in walking than wearable robotics, such as Lokomat [11]. However, more than 1000 Lokomat devices have been purchased and are in use worldwide to restore and improve walking function in patients with neurological disorders. The cost and availability of devices like Lokomat often put therapists and patients under considerable pressure as there is still uncertainty regarding their optimal protocol and appropriate timing of use. Further research on the efficient use of Lokomat-assisted therapy is needed.

Periodic outcome assessments and tracking are fundamental approaches for implementing effective medical practices, which are supported by guidelines issued for stroke rehabilitation [12]. A previous study stated that high-intensity step training applied during inpatient rehabilitation resulted in significantly greater walking and balance outcomes [13, 14]. Although RAGT provides highly intensive and repetitive task-specific training, training programs also need to be individualized and monitored for effective neurorehabilitation. However, data obtained from ongoing RAGT can be best used when organized into a predictive model to help clinicians, patients, and their families with decision-making and planning robot-assisted rehabilitation management as early as possible [15].

During Lokomat training, the body weight (BW) support system and guidance force (GF) provided by the robotic arms assist patients to follow a physiological gait pattern. Later, the support and guidance are accordingly reduced as the patient regains selective motor control. Positive relationships between training parameters and muscle activation were also reported. Specifically, during Lokomat training, reducing BW support and GF was shown to increase gluteus muscle and anterior tibialis muscle activation, while increasing the training speed functions to enhance muscle activation in both legs [16, 17]. However, no predictive model based on Lokomat training data has been established to help identify cost-effective approaches for patients using this system to regain walking function.

Therefore, until the phenotypes for an effective intervention are better clarified, Lokomat-based therapy still relies heavily on therapists’ clinical expertise. We hypothesized that early assessment could accurately predict the effectiveness of RAGT for functional gait recovery based on parameters from the first few sessions. The purpose of this study was two-fold. First, we attempted to determine the most beneficial combination of RAGT parameters in adult patients with neurological diseases in different recovery phases. Second, we anticipated the development of prediction models to assess improvements in the Functional Ambulation Category (FAC) for Lokomat-based therapies in this population. FAC, a commonly used outcome measurement in gait-related studies, is a six-point categorical scale that assesses how much support a patient requires when walking. A score of 0 indicates non-walking, while a score of 4 to 5 indicates an increasingly independent walking ability [18].