Medical technologies, telemedicine and artificial intelligence for neurotrauma and neurorehabilitation

Traumatic brain injury (TBI) is the most common neurological condition, with an estimated incidence of 500–1000 cases per 100 000 people per year and a substantial public health burden, not only as an acute condition leading to death and disability but also for its long-term consequences, including an increased risk of late-onset neurodegeneration [1]. TBI may be secondary to road traffic incidents, which represent the first cause in low-to-middle-income countries, falls in elderly and frail people, criminal offences, and sports trauma, especially in professional contact sports, horse-riding and cycling, and several measures have been proposed to prevent and mitigate its burden [1]. Neurorehabilitation plays a pivotal role in patients with TBI, but only a minority of people with TBI and persisting symptoms and disability are reported to receive long-term in-patient or out-patient rehabilitation [1]. A precision medicine approach that goes beyond the Glasgow Coma Scale and traditional neuroimaging data and incorporates several instrumental and fluid biomarkers has been recommended to improve the assessment and guide the treatment of patients with TBI [1]. In recent years, the availability of high-performance computers, artificial intelligence and machine learning algorithms have shown promise for the development and assessment of different types of biomarkers to improve the detection, monitoring and management of various neurological conditions, focusing on neurodegenerative disorders [2,3]. However, their application to disorders of consciousness, which represent one of the worst complications of TBI, is still in its infancy. The review on ‘Artificial intelligence and machine learning in disorders of consciousness’ by Lee and Laureys (pp. 614–620) presents an overview of artificial intelligence and machine learning technologies to reduce the diagnostic and prognostic uncertainty in patients with severe disorders of consciousness following acquired brain damage and to characterize better patients’ response to novel therapeutic interventions (e.g. zolpidem and transcranial direct current stimulation). Most articles focused on differentiating between unresponsive wakefulness syndrome and minimally conscious state, utilizing artificial intelligence, machine learning and deep learning algorithms to analyse functional neuroimaging and EEG data and assist in clinical decision-making, including the diagnosis, prognosis and treatment. The broader application of thrombolysis and thrombectomy consistently improved the outcome of acute ischemic stroke in terms of the modified Rankin Scale, a global outcome scale that captures disability and represents the gold standard outcome measure for stroke trials. However, the use of a global endpoint to rate may not differentiate between good and poor recovery for specific neurological domains, and many patients still show a marked motor impairment and disability after stroke [4]. Several clinical trials, systematic reviews and meta-analyses reported that noninvasive brain stimulation (NIBS), particularly repetitive transcranial magnetic stimulation (TMS), may improve motor recovery after stroke by modulating excitatory or inhibitory brain circuits [5]. Still, many stroke experts do not recommend repetitive TMS or include it in treatment guidelines because of the lack of uniformity of the evidence and the large variability of reported effect sizes [5]. The review on ‘Noninvasive brain stimulation to improve motor outcomes after stroke’ by Savelon et al. (pp. 621–628) offers an update on recent developments in NIBS applications and techniques for enhancing motor outcomes after stroke, reporting inconsistent effect sizes of interventional NIBS, limited evidence supporting the interhemispheric competition model, and the relevance of metaplasticity and priming approaches to improve the therapeutic efficacy of NIBS. Transcranial temporal interference stimulation and low-intensity focused ultrasound stimulation are emerging NIBS techniques, which may modulate deeper brain structures and hold promise for stroke neurorehabilitation. Additionally, motor-evoked potential obtained with single-pulse TMS and structural and functional reserve biomarkers could stratify patients to tailor NIBS interventions. Cognitive impairment frequently occurs and is one of the most challenging nonmotor symptoms in patients with Parkinson's disease, which is the second most common neurodegenerative disorder, with a 0.3% prevalence in the general population [6]. At variance with the pharmacological advancements in the treatment of motor symptoms of Parkinson's disease, cognitive impairment still represents an unmet need with a relevant impact on patients and caregivers, given the complex neuropathological and neuropharmacological underpinnings and the poor evidence for pharmacological and nonpharmacological interventions [6]. The review on ‘Towards multimodal cognition-based treatment for cognitive impairment in Parkinson's disease: drugs, exercise, noninvasive brain stimulation and technologies’ by Mantovani et al. (pp. 629–637) provides a comprehensive overview of multimodal cognition-based interventions, that is, cognitive stimulation, cognitive training or cognitive rehabilitation, combined with adjunctive drugs, exercise, NIBS and technologies, to treat cognitive impairment in patients with Parkinson's disease. Exercise administered with cognitive training may enhance global, memory, visuospatial and executive functioning; transcranial direct current stimulation combined with cognitive training may improve attention and executive functioning; exergames, semi-immersive virtual reality and telerehabilitation with nonimmersive virtual reality alongside cognitive training may ameliorate global and executive functioning in Parkinson's disease patients. The authors discuss current studies’ limitations and propose recommendations for future trials on multimodal cognition-based interventions for cognitive impairment in Parkinson's disease. Virtual reality allows users to be immersed in an interactive experience that creates the illusion of being physically present in a virtual world, thanks to a head-mounted display and handheld controllers or gloves, which lead to brain and behavioural responses in a controlled laboratory or clinical setting that are analogous to those occurring in the real world [7]. Virtual reality has gained momentum as a versatile research tool to provide personalized rehabilitation in several neurological conditions [7], and research is ongoing on methodological and technological improvements to promote the delivery of virtual reality-based rehabilitation interventions to people with reduced mobility, living in remote regions or who cannot access traditional in-patient or out-patient facilities [8]. The review on ‘Perception and control of a virtual body and immersive virtual reality for rehabilitation’ by Donegan and Sanchez-Vives (pp. 638–644) presents an update on recent advances in using immersive virtual reality to improve bodily perception and motor control in rehabilitation across musculoskeletal and neurological conditions and discusses virtual reality's unique features to address the challenges of traditional approaches, including the potential of the emerging metaverse and the integration of virtual reality with artificial intelligence and brain–computer interface (BCI). Virtual reality shows promise to enhance motivation and adherence, improve range of motion, reduce kinesiophobia in musculoskeletal rehabilitation, and manipulate bodily perception in the neurorehabilitation of upper limb function and gait and balance in patients with stroke and spinal cord injury. Future studies should address limitations in small sample sizes, short intervention durations and variability in virtual reality systems. There is an increasing interest in the use of wearable robotics, which encompasses soft and rigid architectures, single or multidegree-of-freedom configurations, and applications targeting upper or lower limbs for the neurorehabilitation of patients affected by several neurological conditions and to improve autonomy and quality of life in people with different types of disability and impairment [9]. Wearable robotics may be paired with BCI, that is, devices that translate brain activity either invasively or noninvasively into control signals of external devices to enable volitional and intuitive control of exoskeletons in patients with severe brain damage after, for example, TBI or stroke, who do not show sufficient motor function to engage in established rehabilitation strategies, or standard occupational practice [10]. Wearable robotics are moving from the laboratory setting to the clinical practice, and they show promising scores in the users’ satisfaction despite some issues related to ease of adjustment, size and weight, and ease of use [11]. The review on the ‘Assessment of wearable robotics performance in patients with neurological conditions’ by Gerez et al. (pp. 645–654) provides an updated overview of recent wearable robotics in neurorehabilitation, focusing on the assessment of their performance, which is a key point for optimizing the devices and the treatment outcomes. The authors report that, according to the literature, the performance of wearable robots is based mainly on traditional, nontechnological outcomes, that is, clinical scales that are predominantly qualitative and subjective. They also suggest developing standardized metrics that integrate traditional outcomes with the data from embedded sensors to increase the accuracy of assessments and to tailor rehabilitation strategies to the needs of each patient. In conclusion, the reviews of this Trauma and Rehabilitation section report an overview of recent promising developments in medical technologies, robotics, virtual reality, telemedicine and artificial intelligence for patients with neurotrauma and in the neurorehabilitation setting. They also critically discuss several unmet needs, challenges and open issues related to the applications of these technologies and suggest a roadmap for their translation into the real-life clinical setting to improve neurorehabilitation outcomes and lead to personalized treatments. Acknowledgements None. Financial support and sponsorship None. Conflicts of interest There are no conflicts of interest.