Reference | Aims | Type of eye tracker | Eye metrics | Other outcome metrics | Experimental task | Key findings |
---|---|---|---|---|---|---|
Bouwsema et al. (2012) [17] | 1) To provide a description of prosthetic control and performance 2) To relate clinical outcomes to kinematic measures 3) To identify parameters that characterize the skill level of a prosthesis user | Head-mounted (model RK-826PCI, iScan Online, Inc; Dallas, Texas) | Number of fixations and percent fixation | End point kinematics, joint angles, grasp force control and SHAPa | Performed direct and indirect grasping tasks with prosthesis and object manipulation with the non-disabled hand. Objects were solid or compressible. Participants also performed the SHAPa | Two types of gaze behaviours were observed: 1) Visual fixations directed towards the object at the start of the trial and maintained; 2) Visual fixations switched repeatedly between the hand and object. Participants who did not use their prosthesis frequently had a higher total number of fixations, lower percent fixation to the object and higher percent fixation to the hand than frequent myoelectric prosthesis users. SHAPa scores did not correlate to measures of gaze behaviour |
Sobuh et al. (2014) [25] | 1) To characterize the visuomotor behaviours of participants with intact arm function learning to use a prosthesis simulator 2) To compare the visuomotor behaviours of participants with intact arm function and individuals with an upper limb amputation using a prosthesis | Head-mounted (iView Xâ„¢ HED 2, SenseMotoric Instruments GmbH, Tellow, Germany) | Gaze sequence, percent fixation and number of fixations | Movement time and SHAPa | A carton pouring task using the anatomic or the prosthetic hand. The SHAPa was performed as training between testing sessions | When using the prosthesis simulator for carton pouring, gaze was more fixated to the hand in reach and rarely fixated the glass, compared to gaze when using the anatomic hand During manipulation, similar critical areas were focused on regardless of using the anatomic or prosthetic hand. There were significantly greater fixations and a lower SHAPa when using the prosthesis simulator, compared to using the anatomic hand. Training significantly improved SHAP for prosthesis simulator users but had no significant effect on gaze behaviour. Percent fixations, number of fixations during reach, movement times and SHAP scores were similar between subjects with intact arm function and prosthesis users. The number of fixations were higher for prosthesis users during manipulation |
Chadwell et al. (2016) [30] | 1) To assess factors of the prosthesis control chain, including EMG skill and electrode reliability 2) To evaluate performance, including kinematic and gaze patterns and myoelectric prosthesis usage | Head-mounted (Dikablis Professional, Ergoneers) | Number of fixations, percent fixation and percent of look-ahead fixations | Success of task completion, task duration, aperture onset delay, plateau time during reach to grasp, kinematic variability and symmetry of real-world arm use | Participants began by reaching and grasping for a cylinder rotating it 90°, then placing and releasing it into a tube. If over 80% of trials were successful, the same task was repeated with a smaller diameter cylinder. If less than 80% of trials were successful, the cylinder was placed vertically into a vertical tube | Prosthesis User 1 was able to look ahead to the cylinder and tube, whereas Prosthesis User 2 spent most of the time monitoring the hand and cylinder during reach to grasp. Prosthesis User 1 looked ahead of the hand 76% of the time, while Prosthesis User 2 looked at the hand for over 50% of the time |
Zhang et al. (2016) [23] | To compare the cognitive workload of individuals with intact arm function when using a myoelectric prosthesis simulator with direct control or pattern recognition control | Remote (Facelab 5.1, Seeing Machines, Australia) | Percent change in pupil size | Number of clothespins successfully relocated | Clothespin relocation task. Participants moved as many clothespins as possible between the horizontal and vertical bars within a 2-min trial | The pattern recognition group had a greater task performance and lower cognitive load, as shown by a smaller increase in pupil size, than the direct control group. Task performance increased while cognitive workload decreased in later trials |
Raveh et al. (2017) [27] | To evaluate the effects of adding vibrotactile feedback on visual attention and performance of individuals with intact arm function using a myoelectric prosthesis simulator in a dual task paradigm | Remote (GP3 Desktop eye tracker, Gazepoint, Canada) | Percent fixation to the screen | Task completion time and percentage of error in secondary task | Dual task involved using the left hand to toggle arrow keys to navigate a virtual car while grasping activities were performed with the prosthesis simulator. Dual tasks were performed with and without vibration | Adding vibrotactile feedback had no effect on visual attention and task performance in a dual task |
White et al. (2017) [22] | To compare the usability of direct control and pattern recognition control for individuals with intact arm function using a transradial myoelectric prosthesis simulator | Remote (Facelab 5.1, Seeing Machines, Australia) | Number of pupil size increases per second | Number of clothespins successfully relocated and learning percentage | Clothespin relocation task. Participants moved as many clothespins as possible between the horizontal and vertical bars within a 2-min trial | Participants that used pattern recognition had a lower cognitive workload as shown by fewer increases in pupil size, greater task performance and an improved ability to learn compared to direct control. There was also a trend for task performance to increase across trials for both groups |
Chadwell et al. (2018) [18] | 1) To report real-world activity of prosthesis users and participants with intact arm function 2) To investigate whether measures of kinematic and gaze behaviour during a goal-directed task correlate to measures of upper limb activity | Head-mounted (Dikablis Professional, Ergoneers) | Percent fixation | Prosthesis wear time, balance of activity between arms, success of task completion, task duration, delay plateau, reach plateau, acceleration temporal variability | Participants began by reaching and grasping for a cylinder rotating it 90°, then placing and releasing it into a tube. If over 80% of trials were successful, the same task was repeated with a smaller diameter cylinder. If less than 80% of trials were successful, the cylinder was placed vertically into a vertical tube | Prosthesis users relied on their anatomical side to perform daily activities whereas participants with intact arm function relied more on both dominant and non-dominant arms. There were no significant correlations between any measures of everyday use and measures of task performance |
Parr et al. (2018) [24] | To explore the spatial and temporal disruptions to eye-hand coordination during prosthetic hand use in a fine motor task | Head-mounted (Mobile Eye XG, Applied Science Laboratories, Bedford, MA) | Percent fixation, target locking strategy and gaze shifting | Task completion time | Four coins on a board were sequentially picked up from right to left and placed in a jar located in the centre of the board. The task was first performed with the anatomic hand then the prosthetic hand | When using the prosthetic hand, significantly greater visual attention was directed towards the hand and coin and less visual attention to other target areas, when compared to the anatomic hand. In all phases, more time was spent fixating the hand than the target when using the prosthetic hand. There was a significant delay for the eyes to disengage from the current target and shift to the next movement |
Raveh et al. (2018) [26] | To evaluate the effects of adding vibrotactile feedback to myoelectric prostheses on visual attention and performance in a dual task paradigm | Remote (GP3 Desktop eye tracker, Gazepoint, Canada) | Number of fixations to the hand and percent fixation to the screen | Task completion time and percentage of error in secondary task | Dual task involved using the left hand to toggle arrow keys to navigate a virtual car while grasping activities were performed with the prosthesis. Dual tasks were performed with and without vibration | Adding vibrotactile feedback reduced task performance time in a dual task activity performed by myoelectric prosthesis users. There was no effect of adding vibrotactile feedback on gaze behaviour |
Bayani et al. (2019) [15] | To identify gaze strategies that develop implicitly during matched and mismatched limb training during action observation | Head-mounted (Pupil Labs, binocular, Berlin, Germany) | Percent fixation | Movement time, number of errors, type of errors, peak height, peak velocity, peak lateral trunk movement, variability in lateral trunk movement and smoothness | The task involved reaching and grasping for a disc and transporting it over a barrier to be placed in an open slot. Participants watched an instruction video performed either by an actor with a body-powered prosthesis (matched) or with the anatomic limb (mismatched). After watching the video, participants performed the task with the prosthetic hand | In the mismatched group, gaze fixations were directed towards the start and endpoints of the action, whereas for the matched group, gaze was focussed on the path of the prosthesis and the shoulders. With matched action observation, the allocation of gaze shifted from the start and end locations towards monitoring the trajectory of the prosthesis across trials. There was a progressive improvement in motor control in the matched group |
Boser et al. (2019) [16] | To characterize the visuomotor behaviour of transradial and transhumeral body-powered prosthesis users | Head-mounted (Dikablis Professional, Ergoneers) | Percent fixation, target locking strategy, number of fixations and eye latencies | Task completion time, hand distance travelled, hand trajectory variability, number of movement units, peak hand velocity, percent to peak hand velocity, percent to peak hand deceleration, percent to peak grip aperture | A cup transfer task involved moving two cups sequentially across a partition to two target locations then returning the cups to their starting locations. A pasta task involved moving a pasta box from a starting location on the side of the body to a centre shelf, then around a barrier to a second higher shelf, then to its starting location | Transradial body-powered prosthesis users had longer task completion times, increased fixations to their prosthetic hand during reach and transport phases, and movements that were not as smooth compared to individuals with intact arm function. Look-ahead fixations to the drop-off target were within a normative range. Transhumeral body-powered prosthesis users had similar movements as transradial body-powered prosthesis users, however increased fixations to the hand in transport prevented the ability to look ahead to the drop-off target. In the cup transfer task, there were longer fixations to the terminal device during reach than the pasta task |
Hebert et al. (2019) [3] | To determine whether different tasks performed by prosthesis users would result in different visuomotor behaviours | Head-mounted (Dikablis Professional, Ergoneers) | Percent fixation and eye latencies | Movement time and upper body range of motion | A cup transfer task involved moving two cups sequentially across a partition to two target locations then returning the cups to their starting locations. A pasta task involved moving a pasta box from a starting location on the side of the body to a centre shelf, then around a barrier to a second higher shelf, then to its starting location | The cup transfer task required more visual attention to the hand than the pasta task during reach and transport phases, likely due to the risk of spilling the contents of the cup. In both tasks, users had prolonged eye latencies and less fixation on the current target compared to a normative group |
Parr et al. (2019) [21] | 1) To explore the spatial and temporal disruptions to eye-hand coordination during prosthetic hand use in a fine motor task 2) To explore the efficacy of a novel gaze training intervention on prosthetic hand skill learning and retention compared to movement training | Head-mounted (Mobile Eye XG, Applied Science Laboratories, Bedford, MA) | Target locking strategy and gaze shifting | Task completion time, number of errors, alpha power and high alpha connectivity | Experiment 1: picking up a jar filled with water over a barrier to a location on the other side of the board. The task was performed with the anatomic hand first followed by the prosthetic hand Experiment 2: Four coins on a board were sequentially picked up from right to left and placed in a jar located in the centre of the board. A tea making task involved placing a mug on a place mat, adding and then stirring contents with a spoon | With the prosthetic hand, participants focused significantly more on the hand and had a time delay to disengage visual attention in all phases of the jar task. There was also a global decrease in alpha power, indicating increased cortical activation and mental effort. Gaze training increased fixations to the target and speed of gaze shifts, reduced performance time, and improved neural efficiency compared to movement training. These improvements were transferred to a more complex tea-making task. Target locking strategy and faster gaze shifting were significant predictors of T7-Fz connectivity (indicates less conscious control) at retention and delayed retention with gaze training |
Zahabi et al. (2019) [29] | To assess the validity of using a cognitive model to assess the mental workload of upper limb prosthesis use under direct control and pattern recognition control | Remote (Facelab 5.1, Seeing Machines, Australia) | Pupil size | Task completion time and number of clothespins successfully relocated | A single subject performed the clothespin relocation task with two different control modes on two separate days. A cognitive performance model was then constructed to compare the demands of using different control modes | Significantly more clothespins were moved with pattern recognition compared to direct control. Using pattern recognition also resulted in a smaller pupil size, indicating lower cognitive load. The cognitive model indicated that there were fewer cognitive and motor operators with pattern recognition than direct control and no difference in the number of perceptual operators. The model underestimated task completion times |
Kaspersen et al. (2020) [20] | To evaluate the feasibility of using an eye tracker to quantify the sense of agency towards a virtual limb controlled using myoelectric pattern recognition | Head-mounted (Tobii Pro Glasses 2, Tobii AB, Stockholm, Sweden) | Duration of fixations and number of fixations | Reaction time | Four onscreen virtual reality arms were controlled using myoelectric control. Different levels of noise were introduced to randomly reclassify movements to 3 of the 4 virtual arms, making them less controllable. Random arms would flash red at two random time points during each trial and participants were instructed to press a key when they detected a red flash | Two types of gaze behaviours were observed. Participants either fixated on the centre of the screen and used peripheral vision to detect red flashes, or they moved around to fixate on each quadrant. Significantly more time was spent fixating on the most controllable virtual arm, however noise level did not affect the time taken to react to a red flash. Results suggest that visual attention is directed to the virtual arm that provides the best experience of agency to the participant |
Chadwell et al. (2021) [19] | To establish the relative impact of control factors (signal acquisition, signal generation and device response) on user functionality (task performance, kinematics and gaze behaviour) and everyday prosthesis usage | Head-mounted (Dikablis Professional, Ergoneers) | Percent fixation and number of fixations | EMG signal, reaction time, number of undesired activations, electromechanical delay, number of successes, task completion time, reach aperture plateau, movement variability, prosthesis wear time and balance of activity between arms | Participants began by reaching and grasping for a cylinder rotating it 90°, then placing and releasing it into a tube. If over 80% of trials were successful, the same task was repeated with a smaller diameter cylinder. If less than 80% of trials were successful, the cylinder was placed vertically into a vertical tube | A higher number of unwanted EMG activations was significantly correlated to lower success rate, longer task duration, higher temporal kinematic variability, increased fixations to the hand, decreased fixations to grasp critical areas during reach to grasp and increased gaze switches. Longer electromechanical delay was significantly correlated to shorter task duration, shorter length of aperture plateau, decreased fixations to the hand during transport, increased fixations to location critical areas during transport, fewer gaze switches and longer prosthesis wear time |
Marasco et al. (2021) [28] | To quantify the performance of individuals who received targeted sensory and motor reinnnervation using metrics such as visual attention, cognitive demand, fine motor dexterity and ownership | Head-mounted (Pupil Labs, binocular, Berlin, Germany) | Percent fixation | Prosthesis Efficiency and Profitability, Dynamic Prosthesis Incorporation, Grasping Relative Index of Performance, Adaptation rate | A cup transfer task involved moving two cups sequentially across a partition to two target locations then returning the cups to their starting locations. A pasta task involved moving a pasta box from a starting location on the side of the body to a centre shelf, then around a barrier to a second higher shelf, then to its starting location | Providing touch and kinesthetic feedback to prosthesis users with targeted sensory and motor reinnervation reduced fixations to the hand in reach and transport phases, and increased fixations to the next target location. Overall, the integration of bidirectional control allowed users to adopt more natural behaviours |