Introduction

An estimated 7.6 million American adults have had a stroke, and projections show that by 2030, an additional 3.4 million will, a 20% increase in prevalence over the next 10 years [1, 2]. However, recent advancements have driven an age-adjusted decrease in death from stroke and complementary increase in demand for rehabilitation [3]. A given individual’s needs after stroke vary widely as a function of the stroke extent and location. Motor impairments are the most common [4], but post-stroke cognitive impairments have been estimated in as much as half of surviving adults [5, 6] and may include deficits in reasoning, attention [7, 8], memory [9], and language that significantly contribute to a reduced quality of life [10]. Mental health also has been identified as an important mediating factor for rehabilitation success [11, 12].

Post-stroke recovery is impacted by numerous activity-dependent mechanisms including axonal sprouting [13, 1415], dendritic spine elaboration [16, 17], and migration of subventricular stem cells to peri-infarct regions [18, 1920]. Synaptic plasticity is the dominant mechanism for recovery. Thus, the basic principles of behaviorally supported neuroplasticity apply; frequent, rigorous, specific exercises lead to recovery of function [21]. The standard of care for post-stroke rehabilitation remains characterized by task-specific and task-oriented training strategies facilitated by a clinician and deployed for 30–60 min per day for each domain (physical and cognitive-linguistic) in the acute phase and tapering over time as a function of recovery and ongoing access to services.

The goal of physiatric research is maximizing the effectiveness and efficiency of supported recovery. This work can be broadly classified in one of three ways. First is the development and evaluation of activities and strategies to facilitate behavioral modification. Second is identification and refinement of synergistic approaches to behavioral therapy that decrease the threshold for long-term potentiation and depression through direct manipulations like transcranial direct current stimulation (tDCS) [22, 2324] or pharmacological adjuncts [25, 26•]. Finally, there is considerable enthusiasm for the introduction of emerging technology, such as robotics, virtual reality (VR), and gamification for the enhancement of therapy, which offer promising ways of improving rehabilitation adherence. These innovations also allow rehabilitation specialists to introduce and titrate real-world complexity and multifaceted demands, particularly in the inpatient setting. Patients may use these tools in conjunction with other technologies. For example, a patient may engage in a gamified version of a therapy task, meaning it integrates elements like scoring points, rules, puzzles, and competition, typically to increase interest and engagement in the activity’s goal, but instead of moving a joystick or pressing keys, the interaction with the task is electromyographically directed, meaning the electrical activity in the patient’s muscle is the input used to interact with the task.

Here, we will summarize the recent evidence for novel behavioral strategies, synergistic approaches, and technological enhancements across three key domains of function: mobility, cognition, and language. Mobility research has been substantially strengthened by the bench to bedside pipeline. However, there is a relative dearth of cognitive rehabilitation studies targeting attention, executive function, and memory, let alone positive trials. Despite considerable interest, studies targeting post-stroke language rehabilitation, or the treatment of aphasia, are even more niche and, thus, that much more difficult to design and execute. For this reason, we have incorporated both meta-analyses and systematic reviews, which near-ubiquitously note the paucity of well-controlled, sufficiently powered clinical trials, and descriptions of select ongoing trials into our review in order to best reflect the leading edge of physiatry research.

Strength and Movement

Motor rehabilitation was the subject of a recent comprehensive review [27••], which found that the vast majority of novel intervention strategies resulted in no statistically significant improvement in motor outcomes relative to “standard” therapy at the primary endpoint. Null improvements were noted in studies of neuromuscular electrical stimulation, functional strength training, task-oriented training, and modified constraint-induced movement therapy (but see [28] for positive evidence of constraint-induced movement therapy). A review of sensory therapies to improve motor recovery also came up with modest results [29]. Authors noted studies supporting mirror therapy [30] and mental imaging [31, 32] but minimal evidence overall. However, even in the absence of demonstrable improvement over conventional therapy, broadening the diversity of available therapies that are similarly effective provides patients and clinicians with better tools to respond to individual patient needs and preferences, potentially improving overall outcomes. For example, self-rehabilitation of post-stroke motor function is similarly valuable to conventional therapy [33].

Synergistic approaches to motor rehabilitation have included both tDCS [34, 35] and pharmacology [36, 3738] with mixed results, leading authors to highlight the potential value of identifying subgroups to facilitate individualized treatment planning [39]. The strongest evidence for drug therapies comes from serotonergic and dopaminergic drugs [37]; however, no clear pharmacological recommendation has emerged [40]. Intravenous cerebrolysin within 3 days of stroke [41] has been associated with significant upper limb motor improvement and remains a focus of ongoing investigation [42, 43, 4445].

Technological therapy enhancements for motor rehabilitation often are used in concert to create an overall experience for patients that is challenging and motivating. Robot-assisted regimens include passive and active therapy, including electromyograph driven exoskeletons [46]. However, efficacy of robotic implementation varies as a function of the robot and the patient’s needs and may be prohibitively expensive for certain healthcare settings. Recent feasibility study demonstrated the utility of a home-based robotic system for upper limb rehabilitation [47], and the proliferation of robotics into consumer technology may make this a more promising direction for future work. As of this writing, there are a remarkable number of ongoing clinical trials examining enhancements to traditional motor rehabilitation using robotic exoskeletons with (e.g., NeuroExo NCT05374486; NCT04724824; and NCT04599036) and without (e.g., RESTORE NCT04201613; NCT05226988; NCT04054700; and NCT05174676) brain-machine interface. A few trials have demonstrated feasibility of VR [48] alone, VR combined with a gamified therapy activity [49], and VR combined with electrical stimulation and robotics [50] for upper limb motor recovery. However, a systematic review found insufficient evidence to arrive at a conclusion regarding the utility of VR over and above conventional therapy [51]. In contrast, a meta-analysis of 42 trials examining gamified therapy for upper limb rehabilitation found greater, more retained improvements than those from conventional therapy in function, activity, and participation domains [52].

Cognition and Memory

Cognitive rehabilitation strategies most commonly focus on multiple cognitive domains but occasionally target single capacities in isolation. General cognitive protocols have been associated with modest improvements [53, 54, 5556], and additional programs are under investigation (e.g., COMPEX study, NCT04229056). However, a recent Cochrane review failed to find sufficient evidence that cognitive rehabilitation meaningfully improved selective, sustained, or divided attention either immediately or long term [57], with the possible exception of prism adaptation training [58, 59] and visual scanning training [60] for spatial neglect (see section on neglect below). Memory deficits are generally addressed through the use of compensatory strategies and mnemonic aids, sometimes referred to as “cognitive prosthetics.” The proliferation of smartphones and personal computers has revolutionized the kind of support individuals with diverse needs can receive from handheld devices [60, 61] and given rise to new computerized memory training protocols under investigation (e.g., the ASCEND-I study, which combines traditional and gamified elements when targeting working memory; NCT04472351).

tDCS and transcranial magnetic stimulation (TMS) combined with frequent, rigorous, and specific activities have been associated with improvements in cognition. Further, these noninvasive brains stimulation approaches have been found to augment the effects of behavioral therapies targeting executive function [6264]. More recently, studies have been designed to refine our understanding and use of these tools to optimize their effect. For example, an upcoming trial is planned to examine the effect of anodal tDCS combined with computerized cognitive therapy on memory and executive function in individuals with chronic stroke (TIPSCI trial, NCT05195398), while a similar design is ongoing that examines anodal tDCS combined with execution of an n-back task for the treatment of memory and attention after stroke (TRAINS trial, NCT04897334).

Pharmacological interventions continue to be considered for the treatment of inattention, particularly cholinergic therapies, such as rivastigmine [65, 66], though there is a marked absence of double-blind, randomized clinical trials. Presently, clinical trials targeting general cognitive improvement are underway using levodopa (NCT03735901) and maraviroc (NCT04966429), and rolflumilast is under investigation specifically for treating post-stroke memory impairment (NCT04854811).

The emergence of increasingly sophisticated VR technology has generated tremendous excitement [67, 6869]. Two recent meta-analyses of VR in the treatment of cognitive skills found no benefit over standard of care and noted the paucity of adequately powered trials [70, 71], while a third using more liberal inclusion criteria and did find evidence of a benefit of VR on global cognitive measures [72]. Thus, there remains equipoise with regard to the integration of this technology for cognitive benefit. Overwhelmingly, studies have focused on feasibility and failed to generate sufficient information to compare effects with either traditional rehabilitation or spontaneous recovery [73]. As systems become more affordable and commercially available, this is certain to be an area of continued inquiry (e.g., NCT04441177 and NCT05283369).

Language, Speech, and Communication

Post-stroke communication impairments (e.g., aphasia, dysarthria, and apraxia of speech [AOS]) can exist together with cognitive deficits or in isolation (e.g., aphasia, the impairment of language with spared cognitive function). Left hemisphere disorders, such as aphasia, have by far received the most attention in the intervention literature, but disorders associated with right-hemisphere damage (e.g., aprosodia, or the inability to recognize or produce emotional speech) are also common [74] and can be detrimental to an individual’s functional communication. Remediation of communication deficits, to date, is almost exclusively comprised of behavioral interventions, typically delivered face-to-face (although the COVID-19 pandemic did create a monumental shift towards more teletherapy in practice and research). However, with approximately 100 h of SLT needed to significantly improve functional communication [75], interest has turned more and more towards alternative approaches to the recovery of communicative function. These alternative approaches, as with the other common post-stroke impairments, include methods of moderating/enhancing neural activity (e.g., non-invasive brain stimulation or medication) and/or using various technological applications to support recovery. These approaches have not replaced traditional behavioral interventions, but rather typically have been applied in conjunction with them (i.e., as an adjuvant). In all cases of adjuvant approaches discussed in subsequent paragraphs, additional evidence from large RCTs is sorely needed to confirm (or disprove) positive effects on language outcomes.

tDCS and repetitive TMS (rTMS) are the most commonly investigated methods of non-invasive brain stimulation in post-stroke aphasia, with only minimal investigation in motor speech disorders (e.g., [76]). Recent reviews and meta-analyses of tDCS and rTMS have shown accumulating evidence of positive effects on specific language targets, such as naming, but neither can be confidently labeled as effective/efficacious (tDCS: [23, 77, 78, 7980]; rTMS: [81, 82, 8384]). Likewise, additional study of more functional outcomes and individuals in the subacute phase (particularly for tDCS) also is needed. However, evidence is emerging: for tDCS, see Matar et al. [85] for a recent small-N study showing positive results for discourse and functional communication outcomes [85], Sebastian et al. (in review) for positive results of right cerebellar tDCS on functional communication (NCT02901574), or Stockbridge et al. (in review) for positive effects of left anodal tDCS on discourse in an RCT in subacute aphasia (SLISSE trial, NCT02674490). Several ongoing/planned RCTs hopefully will contribute additional evidence to support the efficacy of tDCS (e.g., NCT03773406 and NCT04166513) and rTMS (e.g., REMAP trial, NCT04102228).

There is no current consensus on the best site(s) for stimulation, but most positive effects have been associated with stimulating left-hemisphere perilesional areas and/or inhibiting intact right-hemisphere homologues or the right cerebellum [86]. Specific stimulation sites are generally theoretically determined (e.g., LIFG) but can sometimes be anatomically individualized using resting-state magnetoencephalography (rsMEG; [87]), functional MRI [88], or functional near-infrared spectroscopy (fNIRS; [89, 90]) to identify targets. A comparison of traditional sponge-based electrodes or high-definition electrodes (HD-tDCS) showed similar results [91], although the effect of HD-tDCS was limited to a 2 mA current in a double-blinded RCT [92]. For rTMS, most evidence supports 1 Hz (inhibitory) rTMS at about 90% of the resting motor threshold over right pars triangularis, although one study has shown improvement on different outcomes when inhibiting right posterior STG compared to right IFG [93].

Considering that both tDCS and rTMS aim to enhance neuroplastic mechanisms, it is as yet unclear if one method of stimulation is superior in terms of recovery. There have been very few comparative studies of rTMS versus tDCS and results are contradictory (although this is unsurprising given the significant differences in study design, quality, and stimulation parameters/targets; [94, 95]). Additional study comparing the various delivery parameters for both tDCS and rTMS, individually, as well as comparisons between these and other adjuvants therapies, using comparable paradigms is needed. As a whole, the evidence does show that tDCS and rTMS are safe and relatively easy to use, but they rarely have been implemented into clinical practice. The lack of consensus for the optimal parameters remains a significant barrier to implementation, as well as the lack of access and training for SLPs. Indeed, most SLPs report feeling uncomfortable with tDCS [96].

Another stimulation technique is just beginning to be investigated in aphasia: transcranial alternating current stimulation (tACS; see NCT04375722, NCT05194566). This method of stimulation is used to increase connectivity between two target sites—thus, it is not difficult to see the appeal that this method would have for language recovery considering the breadth of the network. Thus far, only one study examining feasibility has been reported [97].

With regard to pharmacological adjuvants to speech/language recovery, current evidence has been expertly summarized in several recent reviews [25, 26•, 98]. Currently, there are no FDA-approved medications for the treatment of aphasia, but there are a few dozen trials primarily focused on catecholamines (e.g., dopamine and epinephrine) or neurotransmitters such as acetylcholine or serotonin. Few trials have been conducted focused specifically on AOS or dysarthria, although early studies showed conflicting results of dopamine on motor speech, see [99] for review. In general, pharmacological studies have been small and the overall results are mixed but encouraging, warranting further investigation (see [25] for a list of specific drugs recommended for continued investigation). Currently, there are only two ongoing registered drug trials for aphasia: an RCT examining the effect of escitalopram in conjunction with computerized naming therapy (ELISA trial, NCT03843463) and a crossover trial of levetiracetam alone (i.e., not in conjunction with behavioral therapy; NCT00227461). While published drug studies in aphasia focus on identifying positive effects of various drugs on recovery, a recent retrospective study of longitudinal data found no detrimental effects of cholinergic, GABAergic, or dopaminergic medications on language recovery [100].

With technological advances and widespread access, investigation into teletherapy and electronic-based tasks/self-practice is rising considerably. The main premise in pursuing technology-assisted therapy is to facilitate access (by reducing costs and/or logistical barriers like location/transportation) and/or increase the amount of practice an individual can complete, given the large number of hours needed to facilitate learning. Good compliance, ease-of-use, and satisfaction have been reported with these studies [101, 102103], which is encouraging given a primary motivation of increasing total amount of practice time. Results from VR, telerehabilitation, and computer-/tablet-based treatment studies are overwhelmingly positive, but lack of comparison studies and/or adequate control groups/tasks makes it difficult to identify superior protocols or attribute outcomes to the technological component of the treatment [104, 105]. Some studies have shown similar response to remote delivery as in-person, supporting the notion that teletherapy is a viable and effective option [106108]. There are also limited data showing better outcomes with computerized self-practice than some traditional approaches/tasks [109, 110], but any conclusions that one approach is collectively better than the other are inappropriate or, at least, premature. VR applications have been successfully used to create opportunities for real-time or simulated social or task-specific interactions with caregivers or other persons with aphasia [111, 112]. In the case of AOS, besides mode of delivery/practice, computer-based programs have also been used for [bio]visual feedback [113]. In general, most treatment evidence in this area comes from aphasia and AOS, but there is an ongoing trial examining tablet-based practice in dysarthria (NCT05146765).

Behavioral interventions have been, and continue to be, the primary means of intervention in post-stroke communication disorders. The breadth of treatment protocols and targets precludes discussion here, but general trends involve establishing higher levels of evidence for some of the most established protocols (e.g., semantic feature analysis [SFA, see NCT04215952], constraint-induced aphasia therapy (CIAT, [114]), or verb network strengthening treatment [VNeST, NCT05152979]) or to examine optimal dose/frequency/intensity, either generally [115] or for specific protocols (e.g., [116]). The latter question has also prompted investigations of intensive treatment programs where a significant amount of treatment (e.g., 30 h) is provided over a relatively brief period of time (e.g., 2 weeks; NCT04957225, [117, 118]). Such programs report favorable outcomes, including the psychological well-being of participants [119, 120], but the optimal timing of such intensive practice is not known (i.e., acute vs. subacute vs. chronic), especially considering feasibility and tolerance concerns during earlier stages of recovery [121, 122].

There has unfortunately been limited investigation of treatment for communication disorders stemming from right-hemisphere stroke [123] besides cognitive deficits (reported above). Response to a single session of training for affective prosody recognition is reported [124] and an RCT crossover trial treating right-hemisphere communication deficits (aprosodia) is planned (NCT04575909). One previous study reported gains in prosody production with either cognitive-linguistic or imitative therapy [125].

Some researchers have examined the combination of multiple approaches, combining medication, neuromodulation, technology, and/or behavioral language therapy in a single interventional program (combination of medication, tDCS, and behavioral therapy: NCT04134416, [126]), but there is insufficient evidence at this time to draw broad conclusions regarding the benefit of such approaches.

A few other intervention approaches to communication do not quite fit into any of the prior categories but may be of interest to specific providers. These include targeting speech/language and/or psychosocial outcomes through social groups [127], singing groups [128], mental health therapies specifically for persons with communication disorders ([129], NCT04984239), and physical exercise [130].

Swallowing

Traditional dysphagia therapy typically involves behavioral interventions as well as training compensatory strategies (e.g., diet modification or postural adjustments), but recent trends include neurostimulation (both peripheral and central) and biofeedback [131]. Peripheral stimulation includes pharyngeal electrical stimulation (PES) and neuromuscular electrical stimulation (NMES). PES—which is applied intrapharyngeally—has had positive results for tracheotomized individuals with dysphagia post-stroke [132] but mixed results otherwise. Positive outcomes are reported with NMES—which stimulates muscle contractions externally—when paired with traditional therapy [133]. Central stimulations include tDCS and rTMS. For rTMS, best results are associated with high-frequency (i.e., excitatory) rTMS applied to pharyngeal motor cortex—applied bilaterally or contralesionally [134, 135]. Single anode tDCS (applied to ipsilateral pharyngeal motor cortex with the cathode placed suborbitally or on the contralesional pharyngeal motor cortex) offers minimal benefit [134], while bilateral anodal tDCS (anode applied to both hemispheres with cathodes applied to the contralateral suborbital regions) is more promising [136, 137]. As with neurostimulation in the other modalities, optimal parameters are unclear at this time.

Biofeedback for dysphagia typically involves accelerometry, surface electromyography (sEMG), or tongue manometry, and tasks are often gamified to incentivize participation. While individual studies report positive outcomes, a recent review and meta-analysis found that improvements on the behavior targeted did not translate into improved functional swallowing or decreased tube-feeding/oral supplements [138]. Differences in methodological rigor and study parameters may be contributing to the lack of positive findings, thus additional study is needed to determine the benefit (or lack thereof) of biofeedback approaches.

Attention

Neglect is a multifaceted syndrome that can manifest in a variety of ways [139] with most treatment research addressing unilateral visuospatial neglect. While individuals with persistent neglect may be trained to compensate for attentional biases in some contexts (e.g., reading), there is only limited evidence for effective remediation of the impairment [140, 141•]. A lack of consensus stems, in part, from the many differences that exist across treatment study methodologies, participants, outcome measures, etc. Treatment approaches follow other impairments and involve behavioral modification (e.g., prism adaptation, visual scanning, and mirror therapy), adjuvant neuromodulation (e.g., non-invasive brain stimulation and pharmacological treatment), and technology-based interventions (e.g., robotics training), as well as combinations of these approaches.

Prism adaptation is the most common behavioral approach, and the only approach with modest evidence for the treatment of unilateral visuospatial neglect. Prism goggles are worn during therapeutic tasks to shift visual input to the ipsilesional hemispace. While reviews of prism adaptation treatments report positive effects on objective and functional tasks following treatment [58, 59, 142], a meta-analysis of RCTs did not find improvement on subjective ratings or retention of gains beyond 1 month, relative to controls [142]. However, this may be due to the relatively small amount of total practice (i.e., most studies administered only about 5 h total over 2 weeks), differences in the prism parameters (i.e., degree of visual shift), or heterogeneity of recovery stage. Visual scanning training (VST) has yielded mixed results [60, 141•]. Evidence for other behavioral approaches, such as mirror therapy, are scant [141•, 143].

Several recent reviews of non-invasive brain stimulation in conjunction with behavioral interventions such as prism adaptation for the treatment of neglect [144, 145] paint a promising picture for rTMS, tDCS, and neuromuscular vibration (NMV, a type of peripheral sensory stimulation intended to implicitly shift the center of attentional processing; [146149]). Unfortunately, sample sizes, stimulation sites, time-post onset, and other parameters vary such that reliable conclusions regarding the effectiveness of these approaches, especially long-term, remain unclear. Some meta-analyses suggest an advantage for rTMS over tDCS, generally, with excitatory stimulation to the ipsilesional hemisphere and inhibitory stimulation to the contralesional hemisphere yielding equally favorable results [146, 148]. However, a recent RCT showed positive effects, relative to sham, with excitatory tDCS to ipsilesional parietal cortex, but not inhibitory tDCS to contralesional parietal cortex [150]. Evidence for another stimulation technique—continuous theta-burst stimulation (a specific variant of TMS)—is less favorable [148], but perhaps only because there are few studies having investigated it.

Investigation of pharmacological adjuvants follows the same theoretical motivations noted previously, but only a few drug trials were conducted in the last 20 years with regard to neglect. Results are conflicting without positive long-term effects [141•]. The most promising results that may warrant further investigation were for rotigotine (a dopamine agonist; [151, 152]) and guanfacine (a noradrenergic agonist; [153]). We are unaware of any ongoing drug trials for the treatment of unilateral hemispatial neglect.

A single RCT has examined use of robotics for treatment of hemispatial neglect. In that study, a sophisticated robotic toy was placed in neglected space and programmed to interact with the person with neglect. As attention to the device improved, it was moved further and further into neglected space. Participants demonstrated significant improvement on measures of neglect, including performance of activities of daily living, relative to controls, but long-term outcomes were no reported. Another similar study, using an interactive humanoid robot, is ongoing (see NCT05152433). VR applications have limited evidence thus far [141•, 154], but there are ongoing studies to treat neglect in a VR environment (NCT03458611, NCT04651335).

Mental Health

An important caveat to rehabilitation is the frequent and common co-occurrence of post-stroke depression and other neuropsychiatric disorders [155, 156], and their impact on therapy participation and success. For an estimated one in three [156] to five [77] individuals with stroke [157], effective pharmacological and counseling is key to patients experiencing the greatest benefit associated with their rehabilitation therapy. As noted in the prior section, new avenues of research are pursuing the direct treatment of these sequelae for individuals with communication disorders that are notoriously difficult to treat and often excluded from research related to these issues [158].

Conclusions

Rehabilitation after stroke is evolving, with a number of innovative stimulation (sometimes called “electroceutical”), pharmacological, and technological approaches to augment traditional behavioral therapies for the myriad of sequelae of stroke. Most of these augmentations still require large RCTs to demonstrate efficacy and substantial effects on daily function or quality of life.