Stroke rehabilitation is driven by neuroplasticity-based (re-) learning of motor and cognitive abilities with the aim to regain daily-life functioning (55, 59). Neuroplasticity terms the experienced-based adaptation of the structure and function of the nervous system (60, 61). It has been found to be the underlying process of skill learning in animals and humans (62). Since skill (re-) learning is the key goal of stroke rehabilitation, neuroplasticity and (motor) learning principles should be considered for developing interventions for this purpose (55, 56). This may be especially important in the chronic stage after stroke, as in this phase, function regain is no longer supported by the endogenous recovery of the nervous system (63). Furthermore, general training principles should be applied in rehabilitation interventions just as for any training intervention (49, 57, 64). Precisely, objective treatment protocols based on well-reasoned rationales and with defined goals are important for a successful rehabilitation (65). To summarize, we identified three core pillars for the development of our training concept, namely principles for (1) neuroplasticity, (2) (motor) skill learning, and (3) training (Table 2).

Combining these core principles determined the rationale and four “Components” of the PEMOCS concept (Table 2). Building the rationale, the principles “Use it or lose it”, “Use it and improve it”, “Reversibility”, and several more jointly claim that it is crucial for any individual to maintain training for sustaining and/or even improving functioning (49, 55). In line with this, it has repeatedly been shown that chronic stroke survivors can improve functioning with appropriate training interventions (66, 67). To do so, these interventions need to be well reasoned, and personally tailored (50, 51). This led us to the aim of developing a concept that bases on a standard set of rules, which can be applied in a personalized way. Building on this, we elaborated that for applying motor-cognitive training to improve cognitive functions and gait in chronic stroke, four questions (“Components”) need to be addressed; 1) What type of motor-cognitive training is used? 2) At what dosage is the training applied? 3) How is it progressed, and 4) How is it varied to achieve a standard yet personalized intervention schedule?

The Taxonomy for Motor Learning by Gentile (68, 69) was identified as a suitable framework to build the fundament of the PEMOCS concept, and was therefore chosen to be the main parent theory of our Theory Derivation process. Gentile's Taxonomy for Motor Learning is a classification system of different steps in learning motor tasks, and it explicitly considers motor learning principles. It compromises of a two-dimensional structure enabling the classification of motor tasks into different motor skill categories, which represent these different steps of learning (70). The original taxonomy has previously been used in stroke rehabilitation and guided the development of a set of exergames to promote walking ability in chronic stroke survivors (71).

In the Model of Skill Acquisition, three stages of motor skill learning were proposed (72). These stages are (1) an early or cognitive phase, where the trainee first needs to cognitively understand the task they should learn. The goal of the task and possible sub-tasks as well as their sequence are established, involving explicit knowledge. (2) An intermediate or associative phase, where the trainee starts to become more efficient in executing the task, exploring the sub-tasks and improving smoothness and coordination. Finally, (3) a late or autonomous phase, where the trainee becomes proficient in executing the task in varying versions (e.g., different speeds) and can perform it automatically (72–74). The same stages have also been described for cognitive skill learning (75), and find application in Gentile's Taxonomy for Motor Learning (68).

The Cognitive Load Theory states that learning is a process involving a limited amount of working memory, which is needed to process new information, and store it in a comparatively unlimited amount of long-term memory (76, 77). Cognitive load, the extent to which available working memory is demanded by the presented task, is recognised as a determining factor of whether learning is successful or not (77). Therefore, the Cognitive Load Theory suggests that cognitive load should be considered and measured in interventions tailoring (re)learning of tasks.

The Challenge Point Framework states that there exists an optimal level of task difficulty for promoting (motor) learning, and that this optimal level relies on the skill level of the learner (78). Applying that, Guadagnoli et al. suggest to divide task difficulty into nominal, which describes the objective difficulty of a task regardless of the performers skill level or the conditions the task is performed under; and functional, which considers the performer's skill level and the prevailing conditions (78). The Challenge Point Framework claims that, when planning rehabilitation interventions, it is important to consider functional task difficulty instead of only nominal task difficulty (78). This means that rehabilitation interventions must be personalized, which aligns well with recent research guidelines for rehabilitation [(48, 49), see also “1.1 Component 1: Progression Rules”].

The Flow Theory describes a state of “optimal experience”, which explains the motivation and commitment of human beings to activities without obvious external rewards (79). Nine characteristics are proposed to be key attributes of the flow state; challenge-skill balance, action-awareness merging, clear goals, unambiguous feedback, concentration on task at hand, sense of control, loss of self-consciousness, transformation of time, and autotelic experience (79). These can be applied to games including video games, for example in so-called Serious Games or Exergames, which has been described in the GameFlow model (80). The GameFlow model links design of different games to enjoyment of the activity and presents criteria, which can distinguish enjoyable and less enjoyable games (80). Moreover, flow is increasingly measured when investigating neurorehabilitative interventions, as it has been remarked that flow can be beneficial for the success of therapy (81).

The Guided Plasticity Facilitation Model explains how physical and cognitive activities may interact in neuroplasticity. It suggests that physical activity of sufficient intensity facilitates plasticity, while concurrent cognitive activity guides the neuroplastic changes (82, 83). The model claims that both, physical and cognitive activity, are necessary for the induction and retention of neuroplastic changes. Therefore, combined motor-cognitive interventions may specifically trigger neuroplasticity (84). As one of our core principles states that neuroplasticity is central to target in chronic stroke, this corresponds well with our aim to use motor-cognitive exergame interventions in the PEMOCS concept.

FITT-VP stands for the training principles Frequency, Intensity, Time, Type, Volume, and Progression (85, 86). The FITT-VP principles should be applied in any training intervention for any trainee, therefore, also for rehabilitation interventions. The application of the FITT-VP principles also helps to personalize the interventions (85, 86).

The Theoretical Model to Describe Progressions and Regressions for Exercise Rehabilitation describes how a physiotherapy intervention should be structured to achieve successful regain of the target function (87). It claims that the core task of the target function should first be introduced in the simplest version, and then progressed by manipulating internal variables such as speed or range of motion. As soon as the patient is proficient in executing this simple version, an extrinsic component can be added such as a distracting environment or an external force applied. This will lead to a performance reduction compared to the earlier condition, for example in terms of acuity or speed. The task should then be trained under the new condition until proficiency is reached again, before adding a new extrinsic component. By this step-wise introduction of extrinsic components of increasing difficulty, the patient is guided to regain the target function under varying conditions in a standardised yet flexible way (87). This model of progression corresponds well with Gentile's Taxonomy for Motor Learning and the Model for Skill Acquisition described above.

Gentile's Taxonomy for Motor Learning (70) was chosen as fundament of the PEMOCS concept and to build its basic structure. It comprises of a table with two dimensions with two sub-dimensions each (2 × 2 structure, Figure 1). The dimension “Action function” on the x-axis represents the body-internal level of a task and is sub-divided into the dimensions “Body Stability vs. Body Transport”, and “No Object Manipulation vs. Object Manipulation”. “Environmental context” on the y-axis represents the body-external level of a task, and includes the sub-dimensions “Stationary vs. In-motion”, and “No Inter-trial Variability vs. Inter-trial Variability” (70). Combining all sub-dimensions results in 16 motor-skill categories (Figure 1). Task difficulty increases from category 1A at the top left to 4D at the bottom right, which enables step-wise progression through the classification system (Figure 1, (71)). By that, Gentile's Taxonomy for Motor Learning builds a standard framework, through which a learner can move in a personalized manner. The increase in task difficulty by step-wise introduction and then combination of the (sub-) dimensions suits the progression steps suggested in Blanchard's Progression Model for Exercise Rehabilitation (87). It also aligns with the three stages of motor and cognitive skill learning in the Model of Skill Acquisition [see section 3.4.3.4 (72, 75)]. Moreover, besides providing overload by increasing task difficulty, Gentile's Taxonomy for Motor Learning may also provide progression in terms of cardiovascular intensity (55, 57), as introducing “Body Transport”, “Object Manipulation”, “Inter-trial Variability”, and “In-motion” may produce higher cardiovascular load compared to the easier sub-dimensions. This 2 × 2 structure of Gentile's Taxonomy for Motor Learning was adopted for the PEMOCS concept, as it builds a standardised scheme of increasing task difficulty and allows for variability of exercises, which can be applied to individual participants in a personalized manner.

Before starting with the modifications on Gentile's Taxonomy for Motor Learning, we remarked at this point that the components “(1) Training Type” and “(2) Training Dosage” (Table 2) needed to be added to the training concept first. Therefore, the following two sections summarize our findings of the narrative review regarding the questions, “how to choose a type of exergame training?”, and “at what dosage to apply it?”. The further development of Gentile's Taxonomy of Motor Learning, modifications, and further additions were based on these findings.

In the feasibility study being part of the developmental process of the PEMOCS concept, a first draft of the concept with the same motor-cognitive exergame training as described in the application example (Supplementary S1) was implemented and evaluated. This draft was described in (54), where we also identified three main weaknesses of the concept. Following up on this, we describe here what was refined after the feasibility study to address these weaknesses and improve the concept. A more detailed description of the three weaknesses can be found in “Secondary Outcomes—Strengths, and Limitations of the Adapted Taxonomy” in (54).

The PEMOCS concept is intended to serve as a guide for structuring and implementing motor-cognitive training in the rehabilitation of (chronic) stroke. It can be applied in different settings (e.g., inpatient or outpatient rehabilitation, physical therapy, secondary prevention, or similar) and with different equipment (e.g., different VR or exergame systems). As described above, we suggest using a user-centred, purpose-fully designed, and step-based motor-cognitive exergame training to apply it (84, 90, 138). To implement the progression and variability rules based on Gentile's Taxonomy for Motor Learning, an ideal way would be to design a motor-cognitive exergame intervention fitting the different dimensions and sub-dimensions. For example, novel exergames may be designed such as in (71), where six new games were developed specifically meeting the requirements of Gentile's Taxonomy for Motor Learning. The new games require the player to perform balance and stepping tasks with the aim to improve walking in stroke survivors (71). However, it is also possible to assign existing activities to the motor-cognitive skill categories, as we suggest in the “Application example” (Supplementary S1) and (54). To determine the appropriate dosage of the intervention, the above-described recommendations and further practical implications should be considered. For example, the access to the training may play a role, or it should be contemplated if participants need supervision (which would limit the dosage more than if participants can perform the training self-reliantly).

Depending on the setting, the PEMOCS concept may be used to structure a single intervention period or to guide continuous motor-cognitive training in chronic stroke survivors. In in- or outpatient rehabilitation with limited duration, an application similar to the way presented in the application example may be appropriate (Supplementary S1). Such a training period of 12–16 weeks may also be repeated e.g., once or twice a year, if practical reasons (such as patients' schedules) and limited costs [e.g., therapy time, which is paid by the health insurance, (139)] make it more applicable than continuous training. This would consider the training principle of periodization [Table 2, (49)]. Blocked periodization is applied in all types of training schedules, for instance in professional athletes, and was found beneficial as it seems to boost adaption and residual training effects (140). It would also be similar to forced use protocols, such as constraint-induced movement therapy, where intensive training over days or weeks is followed by less intense or resting periods (141). Forced use protocols are effective in improving functioning after stroke (142, 143). Nevertheless, continuously maintaining exercise with the goal to regain functioning and for secondary prevention may be more beneficial for chronic stroke patients (144). For aiding this, exergame training following the PEMOCS concept can be a complement to usual care or help prevent deconditioning when therapy is discontinued (42). Chronic stroke survivors can be guided to maintain a continuous motor-cognitive exercise regimen making use of the progression and variability rules presented here. As long-term continuous face-to-face care is usually not covered by health-care systems (139), an application in a telerehabilitation setting may be more appropriate than the supervised way presented in our application example (Supplementary S1). A telerehabilitative application would mean that patients train self-reliantly at home using technology-based training systems such as exergames, while standing in regular remote contact with an instructing health-care professional (145). For this, the identification of a reliable objective performance parameter (see OP-score, section 3.4.3.3) would be necessary (see Future Directions and Limitations). Moreover, as the long-term use of the same exergames can lead to boredom, we recommend the consideration of further instruments for variability for this application, for instance unlocking new games, adding new levels or changing the exergame system after some time (146).