2. Positive Bike
2.1. System Architecture
The system is constituted by a cycle-ergometer (Cosmed Eurobike 320), a pushing button anchored on the cycle-ergometer handlebars, an Arduino2 board connecting the button to the computer and an Xbox controller. All these components are placed inside a Cave Automatic Virtual Environment (CAVE), a room-sized cube in which the 3D visualization of the virtual environments (VEs) occurs thanks to the combination of four stereoscopic projectors (Full HD 3D UXGA DLP), three rear-projection screens (i.e., the three walls) and one downward-projection screen, all having a projectable area of 266 × 200 cm. A cluster system composed of two HPZ620 Graphics Workstations, mounting Nvidia Quadro K6000 GPU with dedicated Quadro Sync cards, is responsible for the rendering of the four projection surfaces, user tracking and functional logic. CAVE is in fact equipped with a Vicon motion tracking system, with four infrared cameras with 1-megapixel resolution, which allows the tracking of specific reflective markers positioned on target objects and a correct reading of the simulated spaces and distances with a 1:1 scale ratio, thus enhancing the feeling of being immersed in the virtual scene. In order to improve the safety of patients, we decided to implement a cycle-ergometer-based training inside the CAVE, instead of using a treadmill.
In the case of this study, passive markers are used to track the position of the shutter glasses the user has to wear during the exercise in order to change the point-of-view in the VE accordingly. In addition, markers are used to track the heading of the X-box controller; this allows operators to use the X-box controller as a laser-pointing device and to select the interactive contents projected on the screens (e.g., buttons, toggles, etc.) pushing the “A” button on the controller. This input device is not fixed within the system. It is brought in when needed, for example to select the exercise parameters. Conversely, it is brought out of the system when it is no longer useful, for example during the exercise with the cycle-ergometer.
The cycling velocity, as well as the workload, can be read and set thanks to an ad hoc communication protocol developed exploiting the cycle-ergometer software development kit (SDK) provided by the manufacturer; the bike is connected to the computer via a serial cable, as well as the Arduino board.
The VE has been designed and implemented using Unity3D and displayed in the CAVE using MiddleVR for Unity (http://www.middlevr.com/middlevr-for-unity/
). This Unity plug-in provides driver mappings for a variety of existing input devices and accessories—such as Vicon trackers—and delivers abstractions to split functional and graphical logic into a clustered stereoscopic multi display setup, allowing for multi-screens/multi-computers synchronization for higher-resolution VR systems. The hardware setup is shown in Figure 1
2.2. Virtual Environments
The Positive Bike application allows for the accomplishment of dual task training, by providing both cognitive and motor tasks at the same time. The motor exercise will be described later in the article. There is the possibility to choose two different types of cognitive exercise (i.e., using “animals” or “street furniture”) while cycling. In both cases, the cognitive task consists in identifying a series of target objects appearing by the side of the pathway, according to two different criteria. If players are exercising using animals, at the beginning of the exercise, each player is assigned a letter and he/she is asked to identify all the animals whose name starts with that letter (e.g., for the letter “d”, “d
og” is a target, whereas “h
orse” is not). Instead, if the target objects are constituted by street furniture, each player is assigned a color (Figure 2
More in details, Positive Bike VE is composed of three different scenes which are displayed in the following order: (1) login and settings, (2) exercise, and (3) attention test. The login and settings scenes are populated with a 3D graphical user interface (GUI) allowing the therapist either to create a new user or to load the information of an already existing user. Moreover, it permits the setting of the parameters defining the exercise, which are:
Game type: the therapist can set the target typology—and thus define the exercise type—by choosing between animals or street furniture.
Characteristic of the target to select: for animals, the first letter of the animals’ names (C/G/T/S); for street furniture, distinct colors are available (orange/blue /yellow/violet).
Level: two levels of difficulty are available; in level 1, targets appear on the route each 15 s, in level 2, each 10 s.
Cycle-ergometer workload: the operator can set the bike workload selecting among 20/30/40/50 Watt, depending on the patient’s physical condition.
Time: the duration of the exercise, the therapist can select 15 or 20 min.
The selection of the user and of the training parameters are carried out with the X-box controller by the therapist. After the definition of such parameters, he/she can start the exercise.
The exercise scene contains a trail in the park that flows according to the pedals velocity (measured by the cycle-ergometer in revolutions-per-minute, RPM). The path is created thanks to the placement of subsequent nodes on the route, whose interpolation occurs in real-time using quaternion spherical linear interpolation (slerp). Due to the mechanical characteristics of the cycle-ergometer, the user cannot brake or turn intentionally. Being aware of these limitations, the park has been designed to limit any possible desire to deviate from the predefined path. The trail is one and only, and no road forks are present; however, to avoid boredom, elements of the landscape change throughout the exercise, i.e., different species of plants and trees, lakes, buildings, etc. appear on the background. Very slight bends are present to increase the realism of the scenario; prior to the execution of the pilot study, tests on healthy subjects ensured that such curvatures do not induce cyber-sickness due to the expectation of lateral accelerations [60
To elicit an appropriate level of physical effort, participants are instructed to keep their cycling velocity between 55 and 65 RPM: the exercise intensity is adjusted by the therapist who can choose among different ergometer workload values according to the subjects’ capabilities. Two different audio warnings are used to provide the users with feedback when the velocity is too high or too low. In particular, an acute sound indicates that the user is cycling too fast, whereas a grave one signals a too low velocity. Both feedbacks are given in the form of earcons [61
]: this type of auditory display, defined as “abstract and synthetic tones”, were preferred to visual indications because they provide an immediate feedback [62
] without requiring the patients to look away from the path, thus avoiding the possible missing of a target on the way. During the whole exercise, a rustle simulating the cycling on an untarmacked road is reproduced [63
During the cycling, the targets appear randomly, on either the left or right side of the street at a distance of 20 m from the user position, so that the participant has the time to clearly distinguish its appearance and features. Targets’ orientation is random, too. The time elapsing between two subsequent apparitions is driven by the difficulty level selected.
To select a target the user has to push the button positioned on the cycle ergometer handlebar (Figure 1
), before the target gets out of his/her visual field (i.e., it is not displayed on the CAVE lateral wall anymore). Visual feedback is given to the user both for wrong (target becomes red) or right answer (target becomes green). No feedback is given when the user does not press the button, either if the choice is correct (the displayed animal/object is a distractor) or if the target has been missed.
At the end of the exercise, in scene (3), the user is administered an attention question: the application asks the player—via written text—how many targets he/she encountered during the exercise. The user tells the therapist how many targets he/she remembers. The result of this query inserted in the system by the operator, together with session data (date, time, and duration), exercise parameters and user’s performance (number selected/missed targets) are stored in the user’s folder in the form of an XML file.
3. Usability Study
Usability can be defined as the degree to which a specific subject is able to use a given system to achieve specific goals effectively, efficiently, and satisfactorily within a well-defined context of use [64
]. According to this definition, usability is composed by three main factors, all related to the characteristics and the goals of the users and the context of use:
Effectiveness: the possibility for the users to achieve goals;
Efficiency: the effort made by the user to reach the goal;
Satisfaction: what users think about the interaction with the system.
Formative evaluation is a process for the assessment of the usability in order to understand what the usability problems are and suggest developers’ solutions to address work according to expert perspective. In the present study, a formative evaluation was carried out using three validated instruments: The system usability scale (SUS) [65
], an Italian adaptation of the short flow state scale [66
] and a formative evaluation carried out through a semi-structured interview:
SUS is a “quick and easy to use” questionnaire composed by ten items and created by Brooke in the 1996 [65
]. The final score can range from 0, lack of usability, to 100, best usability (for an interpretation of SUS scores, see [67
]). This is a standard scale for the assessment of usability of technological systems and it is easy to use and to understand for the patients.
The short flow state scale [66
] assesses nine key flow dimensions: (1) challenge–skill balance: “I feel I am competent enough to meet the high demands of the situation”; (2) action–awareness merging: “I do things spontaneously and automatically without having to think”; (3) clear goals: “I have a strong sense of what I want to do”; (4) unambiguous feedback: “I have a good idea while I am performing about how well I am doing”; (5) concentration on the task at hand: “I am completely focused on the task at hand”; (6) sense of control: “I have a feeling of total control over what I am doing”; (7) transformation of time: “The way time passes seems to be different from normal”; (8) loss of self-consciousness: “I was not worried about what others may have been thinking of me”; (9) and autotelic experience: “The experience is extremely rewarding”. These characteristics were constructed using the conceptual flow model [68
]. Subjects have to rate the flow experience on a five-point Likert scale ranging from 1 (strongly disagree) to 5 (strongly agree).
The analysis of the transformation of flow could be helpful in evaluating the degree of engagement of patient during the task and, indirectly, the perceived sense of control.
The aim of the “formative evaluations” is to collect information about the usability and interaction from the point of view of the final users. The interview focused on four primary areas: (1) usability; (2) sense of presence; (3) cyber sickness and; (4) expectations. For the first two of these topics, minor themes were identified. In the first case, three systems’ characteristics that give further information about usability were explored; in the second case, three variables that contribute to form the multidimensional construct of sense of presence have been examined in depth:
Sense of presence:
In the Table 1
some exemplificative questions are reported.
The outcome is a description of the main difficulties emerged during the user of application, the impact of the problem on the usability and the practical solutions. The results of the analysis could be used to refine the interaction design.
For the usability assessment, five elderly subjects were recruited—three females and two males. The mean age was 70 (SD 11.70) and the mean years of education (y.o.e.) were 11 (SD 5.61). All the demographic data are reported in the Table 2
Before the session, all participants were given written information about the study and were asked to give written consent to be included. The study received ethical approval from the Ethical Committee of the Istituto Auxologico Italiano.
Subjects with physical problems that prevent riding or with dementia (MMSE < 20) were excluded from the study, no other exclusion criteria were considered. No other cognitive tests have been used because they were considered not related to the usability assessment.
Each subject had to perform the same exercise inside the CAVE. The task required to ride the cycle-ergometer for 15 consecutive minutes inside the virtual environment. The cycle-ergometer workload was set at the minimum level (20W) for all the subjects. The eyeglasses that commanded the visual feedback were hung on the neck of the patients, who had to wear spectacle goggles; this was done to prevent subjects from experiencing cybersickness, such as nausea or dizziness, as while pedaling their head could swing excessively and cause abrupt movements in the virtual environment. Subjects had to keep constant speed during the task. They received an audio feedback if the speed got out of the parameters (Section 2.1
). Everyone used animals beginning with letter “C” as target (camel, dog, kangaroo, horse and deer—respectively cammello, cane, canguro, cavallo, and cervo in Italian) and the lowest difficult level (Level 1).
The results reported were very encouraging and showed that the system had good usability.
The SUS score [57
] was 76.88 (SD = 17.00) and indicated a satisfactory level of usability, according to the questionnaire’s score acceptability ranges, grade scale, and adjective ratings [67
]: indeed, suc score can be included in the third quartile, as showed in the Figure 3
. According to this test, no adaptation of our system would be necessary.
] developed a system exploiting a tablet and allowing chronic obstructive pulmonary disorders (COPD) patients to participate in a videoconference with both other users and the therapists while cycling. In the study, conducted enrolling eight patients, the obtained median value for usability was 81 [70
]. This result belonged to the same acceptability range with respect to the presented system, though it is indeed a little higher. This discrepancy may be explained taking into consideration both the higher simplicity of a tablet-based system with respect to the CAVE environment, and the active participation of the other users in the sessions, which promoted an increased perception of the system’s usability [71
]. State of flow induced by the developed system was comparable with the flow generated by Kinect-based exergames aimed at reducing falls in a sample of adults 65+ [72
] and higher when compared to a Kinect-based dual task intervention developed for balance improvement in patients with Parkinson’s disease [73
] and of an exergame developed with the Wii Fit board dedicated to multiple sclerosis patients [58
Despite the obtained positive results, several issues that could be improved emerged during the formative evaluation. As showed in Table 3
, subjects highlighted some features that, in their opinions, could be modified to improve the quality and the usability of the system.
Most patients reported problems in recognizing animals. In some cases, the problem was related to the dimension of the target. The smaller the animal was, the more patients had trouble recognizing it. To solve this problem, a minimum size for all animals could be established, even if this would involve a less realistic choice. Alternatively, other animals, that have not been inserted in this test because they were difficult to recognize or easily confused with others (for example, pricket, goat, stork; in Italian: cerbiatto, capra, cicogna), could be inserted after a special training with patients to make them familiarize with the animals’ appearance. In other cases, the difficulty was related to the way the animals were presented. If the animal was presented backward some subjects had some problems to recognize it correctly. A simple way to fix this problem is to constrain the rotation of all the animals along the path and make them always face the subject.
Two subjects confused some similar animals, i.e., they called the “horse”, a “zebra” and the “swan”, a “turkey”. A training preceding the exercise with the purpose of familiarizing with animals can be of help in this case too. In addition to this, a general improvement of the quality of the 3D animal models could enhance their recognizability and, thus, the usability of the virtual environment.
A problem related to the discrimination of the audio feedback used for the regulation of the riding speed emerged during the sessions. As said before, an acute tone indicated a too-high speed; conversely, a grave tone indicated a too-low speed. These two earcons overlapped the realistic sound of the bike riding that was provided with the aim of improving the realism, the sense of control and the agency in the environment. One subject had difficulty in differentiating the sounds, especially at the beginning of the exercise. Adding a training phase before the exercise, during which subjects could listen to the different sounds and learn to discriminate them, could be an easy-to-use solution to avoid this problem.
In several occasions, a problem with the button showed up: patients pushed it, but the system did not respond. This technical difficulty could be accommodated by adjusting or replacing the button.
The analysis of flow resulted in a very high mean score: 4.33/5 (SD = 0.84). This indicated that subjects were very involved in the environment and in the task. Csikszentmihalyi [69
] described flow as a sensation that people feel when they act with total involvement. Flow could be also strongly related to wellbeing, because it emerges when good balance between challenge and personal skills is present, that is a situation characterized by high sense of control. As reported in Table 2
, a subject indicated that during the task, she forgot the presence of the examiners. Another subject even perceived that the frequency of appearance of animals had increased, when in reality it was always the same (Table 2
). Analyzing the single dimensions of the scale, it is possible to identify the dimension with the major and minor score. The lowest mean score (3.8) but with the highest SD (1.69) was related to the item “The way time passed seemed to be different from normal” that investigate the “transformation of time”. A high SD indicates a greater variability and lower agreement in the answers among subjects. The totality of subjects responded with the maximum rate (5) at the item “I was not worried about what others may have been thinking of me”. This specific sentence is related to the “loss of self-consciousness” dimension. The subjects forgot the context in which they were during this exercise: they forgot to be in a hospital for a rehabilitation program. This aspect could be a strength of the developed system because it may encourage the patients to be part of the rehabilitation sessions and increase their adherence to the program.
6. Conclusions and Future Works
VR represents a promising technology that, in the near future, can be easily become part of different rehabilitation treatments, as demonstrated by the great number of studies reported in literature [36
]. However, before its introduction in the clinical practice, it is necessary to consider both the pathology-related complications that potential users may have and the usability aspect of the designed system.
A preliminary version of this system was presented in a previous conference paper [74
In this work, an innovative system for motor rehabilitation was presented together with an ad hoc VE developed for the provision of a dual-task exercise to frail patients. The first usability study—conducted on five elderly subjects—with the aim of assessing the system usability and the end users’ satisfaction resulted in a good level of usability, although limited due to small sample.
A strength of our system was the high level of flow showed by the participants associated with high immersion (Q7—Loss of self-consciousness (mean 5/5; SD = 0.00)) and fun during the experience. This indicates that the system is able to engage the subjects more than a classical training program. This is an important factor to consider when designing interactive systems in the medical and rehabilitative field, as greater involvement leads the patient to achieve better results [75
However, different issues related both to software and hardware have been highlighted both by patients and by operators observing the training sessions. These issues will be corrected before the next trial and a training phase, during which the patient is instructed about the tasks and the types of feedback, will be arranged.
Moreover, a limitation of this work is the restricted number of participants involved in the experiment, this problem reduces the generalizability of the results to other subjects. Nevertheless, the homogeneity of the data that emerged from the interviews allows us to hypothesize that many other critical factors with a greater number of subjects would not emerge.
The clinical trial designed to reduce the risk of falling in frail patients is ongoing. The aim of this study is to compare classical and virtual rehabilitation. For the evaluation of this physical rehabilitation program, only motor outcome measurements are considered. The inclusion criteria of the clinical sample are designed to exclude patients with cognitive impairment (MMSE < 26). The clinical trial includes 10 biweekly sections of bicycle and balance exercises of about 30 min for both classical and virtual rehabilitation. The VR program includes another balance task where patients have to avoid rocks flowing to them in a virtual environment. An important part of such clinical trial includes usability and acceptability assessment, since these variables could help us to improve the generalizability and the strength of these preliminary and partial results.
Furthermore, it would also be interesting to compare the performance and the involvement of patients with this system and with another device, particularly with a head mounted display (HMD). An HMD, in fact, could enhance the navigational experience in the VEs and increase the sense of presence, thought it may have the drawback of inducing cyber-sickness [66
]. Future works include also the integration of sensors to monitor the patient physiological status during the training. A heart rate monitor or a breath rate monitor could be easily integrated in the setup to ensure patients’ safety throughout the training sessions and to measure their potential progress during the program.