In recent years, the aging of population in the industrialized countries increased the demand for care services, including neurorehabilitation (1, 2). Then, the overload of healthcare systems and the difficulties in organizing services have required the implementation of new methodologies for rehabilitation (2). COVID-19 has affected rehabilitation processes, especially in neurological patients, harming the quality of life of both patients and their families. To face this unexpected pandemic, new innovative models of rehabilitation service have emerged. At the same time, it led to the increase of non-hospital services to guarantee the continuity of care, thanks to technological innovations (3). Moreover, rapid advances in computing technology have enabled researchers to carry out cognitive training and rehabilitation interventions with the assistance of technology (4).

Cognitive rehabilitation (CR) aims to improve residual neuropsychological capacities through specific strategies based on cognitive models. In particular, the innovative techniques mediated by personal computers (PC) use multimedia and computer resources, through hardware and software systems, to implement cognitive functioning, including attention, memory, problem-solving, language, and executive functions (5–8). Computerized methods are based on repeated training of specific cognitive domains, through the execution of tasks involving specific skills. Most of the tools use audio-video feedback as a motivational stimulus. Furthermore, these tools allow modifying the type, duration, and difficulty of the tasks to adapt the intervention to the individual abilities. The exercises are grouped according to the cognitive domain stimulated and adapt to the patient’s abilities to avoid frustration due to too complex or too simple tasks (5). These devices could offer some therapeutic possibilities for the CR of various neurological diseases (4–8). It has been shown that innovative tools, such as PC-based treatments, could facilitate patient management in the rehabilitation process, allowing continuity of care at home through the telerehabilitation mode (9–13). The tools could support restorative training on cognitive functions thanks to the simulation of different cognitive domains, with positive repercussions on the patient’s motivation (4–8). Various authors have also shown that telerehabilitation can improve various cognitive domains, with results comparable to those of conventional face-to-face rehabilitation (3, 10). Despite the many advantages of PC-based approaches, these devices have also some limitations such as: (a) visual interface limiting their use; (b) access prerequisites (i.e., computer skills); (c) lack of acceptability due to photosensitivity problems; (d) acceptability of devices; (e) reliability (some systems have been validated for the clinic but not remotely, or not validated on some types of the population); (f) availability (some systems are too expensive for a patient to maintain or purchase). However, there are neither clear indications nor warnings for the use of these tools, as rigorous comparisons of technical devices in different neurological populations have never been performed.

In this review, we sought to provide an overall picture of the devices on the market that can be used for CR. Furthermore, a secondary aim is to report on the strengths and weaknesses of the devices employed in inpatient or remote/home settings for continuing the rehabilitation process.

This scoping review was conducted by searching peer-reviewed articles published between 01 June 2010 and 31 December 2022 using the following databases: PubMed, Embase, Cochrane Database, and Web of Science. The aim of the search strategy was twofold: (1) to track progress in the use of software and device-based technology in terms of technological content, human-machine interaction, and cognitive domain training, and (2) to check neurological populations in which the devices and software for cognitive neurorehabilitation are used. To this end, a comprehensive search was carried out using the search terms: (“Cognitive Rehabilitation” OR “Computer-based” OR “Telerehabilitation”) AND (“Stroke” OR “Traumatic Brain Injury” OR “Dementia” OR Multiple Sclerosis” OR “Parkinson” OR “Rehabilitation”). After the removal of the duplicates, all articles were evaluated based on the titles and abstracts. The inclusion criteria were: (i) patients with neurological disease; (ii) a computerized approach applied to cognitive rehabilitation; (iii) English language; and (v) published in a peer-reviewed journal. We excluded articles that described theoretical models, methodological approaches, algorithms, basic technical descriptions, and validation of experimental devices providing no clear translation to clinical practice. Furthermore, we excluded: (i) animal studies; (ii) conference proceedings, or reviews; (iii) studies focusing only on other innovative approaches (such as virtual reality, exergaming, or serious games), (iv) cognitive remediation relating to physiological condition (i.e., the developmental stage or the elderly), (v) study concerning the mobile-app device, which is too far from the traditional CR program.

The list of articles was then refined based on relevance and summarized according to the inclusion/exclusion criteria. Furthermore, to ensure a greater homogeneity in the results, after the removal of duplicates, the articles were evaluated on the basis of the titles and abstracts by two independent researchers (DDB and MGM). These researchers read the full text of articles suitable for the study and performed the data collection to reduce the risk of bias (i.e., language bias; publication bias; time-lag bias). In case of disagreement on the inclusion and exclusion criteria, the final decision was made by two senior investigators (RSC and GM).

Data extraction was performed on 190 articles. Data were considered for the following information: year and type of publication (e.g., clinical studies, pilot study), characteristics of the participants involved in the study, and purpose of the study (Figure 1). After a thorough review of the complete manuscripts, 34 studies articles met the exclusion/exclusion criteria (Table 1). We reported as a primary outcome the one identified by the researchers, for each study, as between-group (in RCT design) or within-subjects difference (for studies with only one group) on the first-level test, and secondary outcome as differences within groups (for RCT) or on second-level tests (for single-group studies). For every study, we selected only significant results adjusted for multiple comparisons.

Although our research in PubMed, Embase, Cochrane Database, and Web of Science has found many technological devices used in CR, only the 10 most cited devices were selected. The information obtained from the selection of studies was organized in two tables. Table 1 reports the list of devices and their main characteristics, as well as the studies and clinical populations on which they were tested. Table 2, indeed, shows what type of study was carried out, how the device was used, and what the results are in terms of treatment efficacy. Most of the selected devices (6 out 11) are supplied in software mode. Then, they can be used by purchasing a stand-alone license, which has a limited duration to the subscription chosen on the manufacturer’s website. Of these, only CogMED (14–19) provides a special license for its usage and a specific training for the online tutor. Three devices (Lumosity, Brain HQ, Brain Gymmer) are either available as PC software or can be installed as an app on tablet/phone devices. Finally, only one represents a telerehabilitation platform (NeuroPersonalTrainer) that allows patients to carry out Hospital and home rehabilitation. In most studies included in Table 2, these devices were tested to evaluate their effectiveness compared to conventional rehabilitation. This review reveals that cognitive training for patients with neurological diseases is covered by several efficient devices that are widely used and validated in the hospital setting, but with few translations in remote applications.

This review aimed to identify suitable technological devices for the CR of chronic neurological patients. Specifically, our literature research has shown how these devices can be used with different neurological pathologies, including stroke, MS, TBI and PD. In detail, it emerges that the clinical population with the most trials is stroke (N = 10) (19–21, 24, 29, 35, 36, 39, 45, 52), followed by MS (N = 9) (18, 22, 25, 30, 38, 40–43), PD (N = 4) (28, 32, 34, 47), and traumatic and acquired brain injury (N = 4, respectively) (14, 17, 23, 26). Only two studies investigating patients with different neurological pathologies were recruited (15, 16), as reported in Table 2. The misrepresentation of RCT studies with such a different neurological population could be intrinsically linked to the difficulty in managing patients affected, e.g., by TBI and dementia. For dementia, there is some evidence that computer-based cognitive rehabilitation may be of help in improving different cognitive domains (57). In particular, the software “GRADIOR” looks promising in the CR field (8, 48–50). Moreover, previous studies have applied computerized approaches using photos of the patient and his/her personal surroundings, with positive results (58, 59). However, these studies were excluded for temporal reasons.

Moreover, most studies reported a statistically significant efficacy of using the PC-based devices in reparative CR. However, only in some cases they were superior to conventional treatments. Training duration, frequency and timing is still unclear. For CogMed, 5 weeks of intervention with each session lasting between 30 and 45 min seems to be the best solution for different populations of patients. However, the efficacy was mainly observed within the group, and not with respect to the control group (14–19). Luminosity™ was mainly used in patients with stroke, but also with MS and TBI, with an intervention duration ranging from 4 to 12 weeks (each session lasting from 20 to 45 min) and with an efficacy higher than that observed for the control group (20–23). Brain HQ needed longer intervention times (12–20 weeks, each session lasting 30–60 min), but with a higher efficacy reported with respect to control intervention for patients with stroke, MS and TBI (24–26). Three studies investigated the use of ERICA for 8–24 weeks (each session 45–60 min), demonstrating significant results only within the experimental group (28–30). CogniPlus (31, 32) and APT (33, 34) have been used for patients with MCI or PD with high variability in the duration of interventions. CoTras (35) and BrainGymmer (36) were both tested in a single study on patients with stroke, the former for a shorter period (4 vs. 12 weeks) and with between group significant differences.

RehaCom was the device more widely tested, especially in patients with MS. The high number of studies increased the variability of the adopted protocols, with a duration of the intervention going from 5 to 15 weeks (each session ranged between 30 and 60 min). However, literature on this device reports solid statistically significant results about its efficacy also when compared to conventional interventions (7, 37–47).

Therefore, we noticed that the importance of training cognitive functions is increasingly evident in the literature, also for facilitating learning processes in motor recovery (4, 60). Indeed, computerized cognitive rehabilitation has proven effective in combination with other methods. A practical example can be the application of acupuncture coupled to transcranial direct current stimulation with computerized cognitive rehabilitation. This method showed good results in cognitive performance in individuals with vascular cognitive impairment (61) and people with stroke (62). A recent study by Shaker et al. (62) demonstrated significant improvement in scores of attention and concentration domains, figural memory, logical reasoning, and reaction times performance. People with cognitive disabilities are treated intensively in the subacute stage of the disease, while unfortunately, they have little access to treatment in the chronic stage. This problem is due to the burden of the Local Health Care Institutions. The underestimation as well as the reduced possibility of effective cognitive training after subacute rehabilitation regards both subjects with central nervous system pathology and those affected by other conditions, such as for example non-CNS cancer (63). This is why new solutions, including telemedicine and home devices/software for cognitive rehabilitation, may be helpful to guarantee the continuity of care. In fact, they should be used when geographical and socio-economic barriers prevent the patient from reaching primary clinics. This will allow each patient to receive monitoring and rehabilitation, through remote devices (3, 64–66).

Moreover, the patient’s perception of the device usability is a key point of rehabilitation. In fact, recent studies have pointed out that the adaptability of technology also includes adapting to patients’ emotions or perceptions. An interesting study by Norman et al. (65) pointed out that perception of a device influences the use of that device itself (37). Nonetheless, this aspect deserves further investigation, as some tools could have high costs and reduce the possibility of customizing the design of the tools. Although in the last period very flexible low-cost proposals have been advanced, also based on smartphones and apps to download for free. We have not explored this field as they are out of the scope of this review. Possible problems concerning the diffusion and use of such devices at home could concern: (i) the absence of a caregiver to supervise the training, especially for patients with greater impairment and with a greater need for therapy; (ii) the lack of experience with technological interfaces and PCs by both patient and caregiver; (iii) the lack of structural technical requirements such as not having a PC or an internet connection. Similar issues have recently been raised by Mantovani et al. (67) concerning the use of Virtual Reality as a home therapy for CR.

In general, it seems that the use of technological devices for CR is promising, but with inconsistencies due to the variations in study design. However, we must bracket the proposal with a caveat. Although these technological devices have features that make them highly adaptive to the patient’s performance. For more severe and subacute subjects they cannot replace conventional CR, in which neuropsychologists and speech therapists play a fundamental role. In fact, their optimal use always remains integrated with conventional CR, or they are part of a rehabilitation process following discharge, to support the patient remotely. However, the protocols of the various studies are very different both in the frequency and the duration of the sessions. This makes it difficult to judge the effectiveness of the tool, so new randomized trials with large samples should be conducted to confirm this aspect. Moreover, it is important that clinicians are familiar with the different devices, in order to facilitate the selection of the appropriate device for the treatment to be performed. Finally, another problem is related to the difficulty of standardizing tests for patients with different neurological pathologies. This implies the need to validate tests for different patients, favoring the continuous updating of devices and tests. Indeed, young subjects, such as those with MS, are more familiar with computerized devices and may require different tests than patients with MCI. Often young patients stop testing because they get bored, or quickly reach the various levels of the tests. On the other hand, patients with dementia and severe cognitive decline may have serious difficulties in using the devices. This could be the main reason why we did not find studies in patients with dementia.

Our review had the ambitious aim of offering an overview of the devices currently in use in clinical practice for the computerized CR of neurological patients. We have collected many studies with the aim of describing the devices and highlighting their strengths and weaknesses. A wide variability among the revised papers was noted in terms of primary as well as secondary outcome measures, even when aiming at measuring the same cognitive domain. This is accompanied by a wide variability also in the duration of treatments, including both session duration and length of rehabilitative period in which a specific device was used (as shown in Table 2). There is the need to standardize assessment and rehabilitative protocols by identifying the key parameter for each device. The inter-rater reliability in the coding and interpretation of these parameters, which in this review cannot be performed given the wide variability among the studies. Thus, this work has limitations. Unlike validation studies, it is not possible to operationalize and define the key parameters being analyzed in the identified literature and then demonstrate inter-rater reliability in the coding or interpretation of each of the defined parameters. The scientific literature on this topic is very varied: different devices are used, for different types of patients, administering a different amount of therapies/duration. Further meta-analysis reviews are needed to fulfill this purpose. In the near future, various factors can consolidate and improve the possibility of carrying out cognitive therapy using software and platforms at home. They include: (i) better accessibility (in terms of lower costs and greater geographical coverage), (ii) higher attention to the chronic and territorial phase of neurorehabilitation and (iii) a growing sensitivity to the possibility of ensuring a better quality of life for brain injury survivors.

With this in mind and considering the aforementioned limitations, PC based approaches could be valuable complementary tools to improve cognitive function and partly guarantee the continuity of care in neurological patients.

MM: Conceptualization, Formal analysis, Methodology, Writing – original draft. DB: Formal analysis, Investigation, Methodology, Writing – original draft. RC: Investigation, Methodology, Writing – original draft, Writing – review & editing. IC: Supervision, Validation, Writing – review & editing. AC: Validation, Writing – review & editing, Supervision. PT: Supervision, Visualization, Writing – review & editing. FI: Software, Validation, Writing – review & editing. SP: Software, Writing – review & editing. GA: Software, Validation, Writing – review & editing. GM: Methodology, Visualization, Writing – review & editing. MI: Supervision, Visualization, Writing – review & editing.