Comparison Between Virtual and Real Settings

Initial studies on the assessment of people with TBI have sought to measure whether performance was the same between VR and the real environment. Among earlier studies, Christiansen and Zhang (Christiansen et al., 1998; Zhang et al., 2003) introduced measures to compare virtual performance and real task performance when preparing a simple meal. The virtual environment provided a simulated kitchen environment on a computer screen with which patients can interact using the keyboard and the regular mouse. The main objective of these studies was to assess the impact of cognitive disorders after TBI on meal preparation and to compare virtual and real performance in this task. Task analysis was based on a step-by-step approach to fulfill the task and on the patients’ abilities to spontaneously undertake every sequence, or with the need of cues. Totally, 54 patients with severe TBI were included in the study. The authors’ findings showed strong correlations between performance result in the meal preparation and between real and virtual settings. Furthermore, they found correlations between the virtual meal preparation task and neuropsychological tests, highlighting that cognitive processes were associated with this kind of simple task.

In a recent study to compare virtual and real performance, our team compared tw'o groups of people with moderate to severe TBI in a route-learning task (Sorita et al., 2013). The virtual district was simulated from a real urban district just next to the hospital of Bordeaux where the participants were inpatients. The virtual setting was semi-immersive with a video projection displayed on a large screen (Figure 7.1). Patients were seated in front of the screen and moved themselves in the virtual district using a joystick. The learning procedure was based on the route learning task (Barrash et al., 2000). The route distance was approximately 800 meters. The assessor showed the route once and subjects had to repeat the same route twice; immediately and after a delay of one day. After doing the route three times, subjects had to complete three spatial cognitive tasks, two map-like tasks, and one chronological arrangement of pictures from the route. Learning route scores did not show any differences between real or virtual learning settings. Furthermore, the learning curves were strictly parallel. The only task that differed between the real setting and VR was the picture arrangement task. We suggested that this discrepancy could be due to the lack of sensorimotor input during the virtual task compared to the real task, sensorimotor inputs being usually involved in these kinds of spatial representations.

According to the results of comparisons between performances carried out in virtual or real settings, a key question is whether the same cerebral networks are implicated in both environments. Slight differences have been shown when comparing performance in real and virtual settings in studies using functional magnetic resonance imaging (IRMf). Part of the cerebral networks involved is the same between settings, but virtual tasks involve specific cerebral networks due to the use of technology (see for instance Mellet et al., 2010). One suggestion from these studies is the importance of familiarizing subjects to the specific use of VR technology when doing a task, to reduce the bias of cognitive workload that could alter the task progress (Besnard et al., 2016).

Comparison Between People After TBI and Healthy Control Subjects

Regarding the comparison between people after TBI and Healthy control subjects (HC), a study published in the early 2000s assessed a group of patients with dysexecutive syndromes and notably with planning disorders (McGeorge et al., 2001). The virtual environment was displayed on a 17" computer screen and subjects interacted with a standard mouse. The subjects’ performance was compared with the same task in a real setting using a video analysis. The authors designed an original task. Subjects had to plan and perform several subtasks that were part of an office workflow. Principles of the MET (Shallice and Burgess) were implemented to design this task. A TBI group of five patients was compared with a HC group. The comparison showed the power of the task to discriminate the quality of performance between groups. The TBI group performance was significantly lower than the HC performance due to dysexecutive syndrome. Performance in the virtual environment was significantly correlated with performance in the real environment for both groups.

Other studies focusing on the assessment of cognitive impact on behavioral performance have been published subsequently, with tasks designed on the basis of neuropsychological models. The main objective was to study the psychometric properties of virtual tasks and their sensitivity when comparing TBI and HC groups. For instance, Renison et al. (2012) included 20 participants in a post TBI group, with onset of TBI more than one year, and a HC group. Executive disorders were characteristic in the TBI group. The authors wanted to compare the performance of the groups in a real and virtually simulated library task. In this task subjects had to follow predefined rules. Instructions were displayed on the screen depicting the task to be performed. Tasks were designed on the basis of seven theoretical dimensions of executive functioning. The virtual environment was not immersive, using a computer screen and a gamepad. A conventional battery of neuropsychological paper and pencil tests was also administered, in addition to a questionnaire aiming to measure dysexecutive syndrome in daily life. Lastly, the virtual task was compared with a similar real library task with the purpose of assessing the EV of the virtual task. Results showed significant differences between groups, whether the task was done in VR or in the real setting. In the virtual task, four sub scores out of seven summarizing the executive functioning components were significantly different between groups. Components of prospective memory, inhibition capacity and multitasking were significantly different between groups. For the two groups, performance scores in real and virtual tasks were highly correlated with paper and pencil neuropsychological tests as well as for six of the seven executive functioning scores. Furthermore, the total score and three sub scores of the virtual task showed significant correlations with the daily living questionnaire. In this study, the authors demonstrated the EV of the virtual library task with a construct that reflects both the underlying neuropsychological processes implicated in the task and the behavior in daily life.

Correlations of virtual environments with measures aiming to capture daily living functioning are not sufficient in themselves. The development of virtual tasks also has to show convergent validity with neuropsychological tests intended to measure underlying injured cognitive processes. In a recent and interesting study, Besnard et al. (2016) compared the performance of TBI and HC groups in a virtual task of making a coffee in virtual and real kitchens. The TBI group included 19 subjects in a post-acute to chronic phase after TBI. The HC group included 24 volunteers matched with the TBI group for years of education and gender. All the participants received a familiarization training in the virtual environment. VR was non-immersive, using a 17" laptop screen. VR performance indicators were the time to complete the task, percentage of completed sequences with or without errors, the total number of errors, with a qualitative identification of error-type classified into two main categories of omission errors (to forget an action), and commission errors (to do a non-expected action). These categories of error are commonly reported in the literature as a consequence of dysexecutive syndrome (Poncet et al., 2017). Comparison between HC and TBI groups showed significant differences between the two modalities, real and virtual, in the five task performance indicators. In the HC group, time to completion and total of errors correlated significantly between the two modalities. Interestingly, the authors noted that despite correlations, the number of errors as well as the time to completion was significantly higher in the virtual kitchen than in real kitchen for both groups. They suggested that VR was more sensitive to action impairment than the real environment. This discrepancy was due to the greater additional attentional demand in VR for persons with TBI, who have less resources than healthy subjects. Another hypothesis was that action schemas were not as intuitive as in the real setting, leading both groups to have a performance discrepancy. To conclude, the authors suggested that, despite the demonstrated EV of the virtual task, this particular task was more cognitively demanding than the same task in a real setting.

The development of cognitive assessments using virtual-simulated tasks can also be used to screen for specific neuropsychological diseases. For example, PM is a specific type of memory that can be particularly affected after TBI. PM can be described as memory for events occurring in the future and is therefore very important for optimal functioning in daily life. VR offers interesting potential in relation to the evaluation of PM. Canty et al. designed a shopping task test based on a cognitive model of PM. The task was based in a virtual shopping mall and was designed to be sensitive to PM. The authors tested the EV of VRST to demonstrate its close theoretical model construct and at the same time, they were interested in whether it could predict the impact of PM disorder on daily life functioning. Twenty-five patients with severe TBI. with an onset of less than two years were included in the study. The HC group consisted of 24 healthy adults. The VRST was presented on a laptop computer and commands were carried out using the keyboard keys. The shopping task involved purchasing 12 items in a prespecified chronological order, in a mall consisting of 20 shops. During the shopping task, subjects had to carry out three time-based PM tasks and three event-based PM tasks. The total duration of the task was 14 minutes overall. The results showed significantly lower scores for the TBI group compared to HC group for the overall scores on the evaluation battery, and notably for the time-based PM and event-based PM tasks. Moderate correlations were found between PM tasks and independent life skills, demonstrating the relative importance of PM in daily life. This study concluded that the virtual PM task was of clinical interest in a controlled virtual setting compared to real-life setting, but also was more sensitive than traditional PM assessment tasks. In addition, measures from the VR tasks were the strongest predictors of psychosocial functioning.

< Prev   CONTENTS   Source   Next >