After the visual and automatic inspection of EEG tracks, four cases in the O and 2 cases in the Y groups were discarded because of the low quality of the recordings; so the final evaluation was carried out in 4 F and 6 M in the O group (mean age 65.4 ± 8.23) and 5 F and 7 M in the Y group (mean age 25 ± 3.8).
Virtual reality method: home environment reconstruction
The requalification started with the acquisition of the real environmental parameters, such as room dimension and their shape. The starting real environment was a laboratory of the Polytechnic of Bari. The rooms, after they were modeled in Virtual Reality, were furnished as a living environment and illuminated with different spot lights, by means of domestic remote control. The steps needed to create a virtual model were:
Acquisition of real environmental measurements through a high precision laser (Leica 3D Disto);
The points acquired by the laser meter were exported to a CAD file, processed using Autodesk Inventor, from which a rough 3D model of the environment was created;
The rough model was enhanced with furniture, so the model was made look like a home environment containing two corridors, three rooms (living room, bedroom, and a kitchen) and a bathroom, positioned in the corridors.
From the basic model (Fig. 1a), other two different models were created, where each one contained the bathroom,—the target place—in a different position, such as in the living room (Fig. 1b), and in the bedroom (Fig. 1c). Starting from each object, three different scenes were generated, each containing one of the three models mentioned above (Fig. 2).
In order to immerse the user in the VR scenes, an Oculus Rift DK2 was used, which is a virtual reality head-mounted display headset developed by Oculus VR. For the purpose of this project, when the user wore the device, he saw the virtual scene through which the visual stimulation occurred. The binocular vision describes the way in which we see two views of the world simultaneously; the view from each eye is slightly different and our brain combines them into a single three-dimensional stereoscopic image, an experience known as stereopsis. The Oculus Rift presents two images, one for each eye, generated by two virtual cameras separated by a short distance (Fig. 3).
The Oculus Rift DK2 employs an OLED panel for each eye, with a resolution of 960 × 1080 and a refresh rate of 75 Hz (it globally refreshes, rather than scanning out in lines). The panels have low persistence, meaning that they only display an image for 2 ms of each frame. This combination of higher refresh rate, global refresh, and low persistence means that the user experiences none of the motion blurring or judders that are experienced on a regular monitor. It uses high-quality lenses to allow for a wide field of view. The separation of the lenses is adjustable by a dial on the bottom of the device, in order to accommodate a wide range of interpupillary distances. In order to work, the Rift was connected by a cable to a PC equipped with a powerful GPU, at least, equivalent to an NVIDIA GeForce GTX 970 or AMD R9 290, and a CPU, at least, equivalent to Intel i5-4590.
In addition to the 19 standard positions of the international 10–20 system, 38 additional electrodes were placed on the x, y, and z coordinates provided by the Advanced Source Analysis (ASA) software (ASA version 4.8.1; ANT Software, Enschede, The Netherlands;httpHYPERLINK “http://www.ant-neuro.com/”://www.ant-neuro.com; http://www.ant-neuro.com). The reference electrode was placed on the linked mastoids, the ground electrode was in Fpz and two electrodes were placed above and below the right eyebrow for electrooculogram (EOG) recording. The impedance was kept at 5 kms or less. The EEG and EOG signals were amplified with a bandpass filter of 0.5–80 Hz digitized at 250 Hz and stored on a biopotential analyzer (Micromed System Plus; Micromed, Mogliano Veneto, Italy; http://www.micromed-it.com).
During the sessions, the participants sat on a reclining lounge chair located in a sound-attenuated, electrically shielded and dimly lit room. At the start of the sessions, they were asked to avoid blinking as much as possible. The study was approved by the Ethic Committee of Bari Policlinico General Hospital. All subjects were informed about the procedure and signed an informed consent. The trial was included into the Rescap Apulian Living Lab project (http://livinglabs.regione.puglia.it/).
The P3b paradigm was designed in order to test the best features of a target place to be recognized in a virtual ambient reproducing a real house. We selected the bathroom door as target stimulus, for its presence in all houses and health facilities, where an elderly subject could have difficulty in moving toward this room essential for daily living abilities. In the virtual environment, that simulates a home, all rooms that were different from the bathroom were identified as frequent stimuli (F) while the rare stimulus was assigned to the bathroom room itself. To differentiate the target stimulus from the frequent ones, and to understand the better features for target room doors recognition, bathroom doors were semi-opened, in order to allow the recognition of the typical furniture. In addition, different color and luminous conditions, easily obtained in the real ambient from a remote home automation control, were applied. Bathroom doors were illuminated in white, like the other non-target doors (W), or colored with a green (G) or red spotlight (Fig. 2). Subjects were informed that they would perform a virtual walking through an apartment, looking for the bathroom door, and were requested to press a button as soon as the bathroom door appeared in the field of view. To make the experiment easier, navigation inside the environment was automatically controlled and with a fixed duration; in this way, rare and frequent stimuli were proposed in succession and in a controlled number. In order to examine the best bathroom position for being recognized within the virtual house, three different Virtual Environments (VE) were depicted, as the bathroom was designed in the aisle (A), living room (L) and bedroom (B). Each VE constitutes a single block (Fig. 1), where each block included 30 rare stimuli (10 for each color spotlight) and 120 frequent ones (all other rooms doors). Thanks to the anchors inserted in VE, the system computed random virtual paths that, alternatively, started from the bathroom, end up to another place of the house and vice versa, in order to have a rare stimulus followed by a certain number of frequent ones.
During the automated navigation in VE, whenever an open door entered the field of view of the virtual camera, a trigger was launched to the EEG recording system according to the door type; in particular, bathroom door throws rare trigger, with different values depending on the spotlight colors, while the doors of the remaining three rooms launch the same frequent trigger value.
During the experiment, a virtual agent, equipped with a virtual camera that reproduces its visualization on the user’s head-mounted display, followed the route. In order to allow the health technicians to supervise the experiment evolution, an operating mode of the application was developed. In this modality, the operator saw the same virtual environment visualization of the patient and other information related to acquisition protocol. A monitor, used exclusively by the operator, was in fact employed. In operator display, trigger information showed which type of stimulus was visible in the field of view and, accordingly, which trigger code was sent to EEG recording system. Other information provided to the operator were the number of stimuli sequence block, and the next destination, e.g. “bathroom” and “other place” (Fig. 4).
We have also to outline that the color modifications reproduced in VR were realized in the real ambient, by means of different lights activated by a remote control.
In order to test the reliability of the P3b obtained in the virtual ambient, subjects were also submitted to a standard oddball visual paradigm presented on a monitor in a fixed position, with the same percent rate of the target and no target stimuli. This recording was performed before or after the VR paradigm, in a randomized inter-subject way.
We measured the positive component (P3b) to detect the neural activity related to the target stimulus. Averaged epochs were 1100 ms, beginning 100 ms before stimulus onset. Waveforms were averaged off-line after visual inspection, such that trials in which the EEG or EOG exceeded ±65 uV were rejected by the automated procedure provided for artifacts by the ASA-ANT software. In addition, the Independent Component Analysis (ICA), provided by ASA software, was applied to further subtract from EEG tracks the features of the visually inspected EEG artifacts and EOG. We evaluated P3b for each presented target stimuli color (W, R, and G) in each block (A, L, B). We averaged responses to W, R, G and F stimuli in the three virtual environments (A, L, and B). Each averaged response was visually inspected and compared to the responses obtained in the fixed-display condition. Only subjects with at least 7 artifacts free responses for each presented target stimuli color in each block, who in addition did not show substantial modification of P3b latency and amplitude with respect to the standard visual oddball paradigm, evaluated by visual inspection, were admitted to the final analysis. Absolute latencies and amplitudes were automatically computed by ASA software at the highest positive peak in the time range 300–600 ms after stimuli delivering. The amplitude of each wave was measured from the baseline. A grand average (GA) of ERP was performed in the two groups:
Fourteen subjects 7 M and 7 F in the 60–80 years age range (O—old group).
Fourteen subjects 7 M 7 F in the 20–30 years age range (Y—young group).
All subjects were selected from hospital and University staff or were familiars of these. Exclusion criteria were the absence of history or objective signs of any general medical, neurological and psychiatric disorders, including a Mini Mental State Evaluation Score (Folstein et al. 1975) <26 for subjects in the old group, and central nervous system acting drugs taking. The inclusion criterion was a scholar age of at least 13 years. Patients under treatment with mild hypertension were admitted. The Mini-Mental test score in older patients was 27.32 ± 1.27, with a mean secular age for older patients of 14.57 ± 2.20 and for young participants 15.14 ± 2.07 years (ANOVA F 0.44, n.s.). Visual acuity was also measured by an oculist, and only subjects with normal visual function in natural conditions are obtained by contact lens or glasses (which were allowed during the task) were admitted.
In order to evaluate the changes of P3b amplitude with regard to the color of the target stimulus, the different virtual environment (aim 1) and the two age groups (aim 2), we employed MANOVA test. In the first step analysis, we introduced the amplitudes of the artifact-free maximal positivity in the time range 300–600 music recorded over the employed 57 EEG derivations as variables in the MANOVA test, where groups Young versus Old and target stimuli W versus R versus G v’s F were the main factors, also evaluating the interaction between groups × target stimuli. In a second step analysis, we introduced the same variables as in the first step, considering the effect of Virtual Environments (A vs. L vs. B) on the P3b by target stimuli, and the interaction VE × groups. The P3b latencies were evaluated by one-way ANOVA, considering Pz derivation as variable and the group as a factor. A posthoc Bonferroni test was run out in single groups to test significant differences of P3b amplitudes and latencies among target stimuli and Virtual Environments. Scalp Maps reporting the Grand Average of the P3b in single groups, and Statistical Probability Maps (SPM) reporting the results of Bonferroni test, were constructed according to the tri-dimensional scalp model provided by ASA software.