Brain Injury Professional

The official publication of the International Brain Injury Association and the North American Brain Injury Society

Eye Tracking Technology: A Primer

By Jacqueline Theis, OD, FAAO

Though seemingly simple, eye movements, also known as oculomotor function, require an expansive, multifaceted neural network to seamlessly provide visual information to the brain and allow for cognitive and physical interaction with our environment. These neural networks involve every lobe of the brain, brainstem, cerebellum, thalamus, basal ganglia, cranial nerves, and visual tracts. The oculomotor system also interacts with the vestibular and cervical systems to ensure proper reflexive movement and proprioception when the head and body is in motion. Due to its extensive neuroanatomy, the oculomotor system is highly vulnerable to neurologic injury.

Oculomotor impairment is common in acute,1 subacute,2 and chronic3,4 time periods after traumatic brain injury (TBI), and the presence of new oculomotor dysfunction post-TBI may be associated with worse outcomes and protracted recovery.5,6 Early detection allows for improved management of patient expectations and impacts the treatment timeline by expediting the referral process. Thus, oculomotor assessment is imperative to the diagnosis and management of TBI.7 The emergence of eye tracking technology has revolutionized oculomotor assessment and has the potential to be an accurate, sensitive, and objective biomarker for neurologic function and brain injury.8,9

There are numerous benefits of utilizing eye tracking technology in clinical practice. First, eye trackers are often compact, portable devices that can provide testing in a variety of clinical and non-clinical spaces. Theoretically, they document more objective and accurate evaluations than a clinician, providing a more sensitive baseline, and can quantify the dynamic components of eye motion, allowing for more accurate monitoring of recovery. While eye movements can be subjectively manipulated based on cognitive effort and attention (i.e., the patient decides if they want to look at a target or not), overall, oculomotor metrics like velocity, amplitude, and gaze path are reflexive, providing an objective diagnostic measure. Eye trackers are non-invasive, and can be performed quickly, in stark contrast to neuroimaging. While offering the potential to generate a revenue stream for providers as a reimbursable procedure, eye tracking is still cost efficient for both the patient and insurance provider. Finally, there is immediate patient satisfaction. TBI, particularly concussion, is an invisible injury which can cause added psychological distress to the patient. Oculomotor dysfunction detected by eye tracking can provide symptom validation for the patient and confirmation of TBI diagnosis when combined with clinical examination.

Eye tracking innovates TBI assessment and management, but buyer beware; accurate oculomotor metrics can be elusive, and current technology is still in the experimental phase and is not as reliable and insightful as a standalone assessment as clinicians may hope. The ultimate potential for value of an eye tracker is to see eye motion that cannot be physically seen by a clinician, more precisely with quantitative data, and analyze pattern differences for diagnosis and management. To identify the best eye tracker for your practice, you need to understand the system specifications that differentiate the products on the market, as the “perfect” eye tracker does not yet exist.

Eye Movement

All types of eye movements are susceptible to brain injury, and every type and direction (vertical vs horizontal) of eye movement has a different neuropathophysiology. Thus, if you want to use oculomotor assessment as a biomarker you should evaluate every eye movement in each direction. Unfortunately, there is yet to be a device that can reliably assess all eye movements in one sitting, so you may need multiple devices to provide a comprehensive evaluation.

Data Validity and Reliability

The smallest and fastest eye movement that can be detected is dependent upon the spatial and temporal resolution, accuracy of the eye tracker, and the reliability of its data. This is dependent upon multiple factors such as imaging quality, signal to noise ratio, and the algorithms used for detecting and tracking the eye.

In eye tracking, validity implies that the eye motion data measured accurately corresponds to the actual position/movement pattern of the eye. Devices list gaze accuracy in degrees of visual angle, as a measure of the spatial distance between the true and measured gaze direction. Eye movements vary in size and speed and depending on which eye movement you want to track the accuracy of the machine can be crucial. Keep in mind, that accuracy is assessed on artificial eyes, but can vary clinically because it is dependent on the individual’s eye characteristics and the testing environment.10,11,12  This is why many devices require calibration before every patient use.

Precision, on the other hand, measures the consistency and repeatability of gaze point estimation if the eye movement being tracked was constant. The lower the precision of a device, the less reliable the data. It is suggested that precision should be below 0.05° to accurately measure fixations and saccades and lower than 0.03° to accurately measure microsaccades.12

Head Movement

Head movement impacts eye movement. Eye trackers can be a wearable head-mounted device like a helmet or spectacle frame, a free space device like a cellphone, or a more stable desktop device. While head-mounted and head-unsupported free space devices can provide real-world eye tracking outside of the clinic setting – on the sideline perhaps – the precise fit of the device and movement of the head, device or face due to facial expressions or speaking, can impair data quality.13 Wearable and head-unsupported eye trackers use software algorithms to compensate for head movement, but these algorithms still have a limit of how much the head can tolerably move for accurate data acquisition. Thus, the advantage of a desktop device with head stabilization with a chin/forehead rest has less head movement and offers increased accuracy, repeatability, and spatial resolution.14

Sampling Rates

Eye trackers use cameras to take multiple images of the eye position over time, and then software determines the eye movement pattern based upon algorithmically predicted eye position and sampling rate. Sampling rate (frequency) indicates the number of times per second (Hz) the position of the eyes is registered by the tracker. The higher the sampling rate, the more likely the eye tracker can estimate the true path of the eye as it moves, and the higher the precision. The lower the sampling rate, the greater the time interval between eye position detection, allowing for increased under-representation of eye motion abnormalities and possible misdiagnosis of “normal.” A higher sampling frequency will more accurately detect the gaze path but is more expensive as it requires more advanced cameras, illumination, and data storage. Every eye movement has a different frequency, and according to the Nyquist-Shannon sampling theory, the sampling rate should be at least 2x the speed of the eye movement you want to capture.15 So to measure a saccade you need a sampling frequency higher than 200-300Hz to accurately calculate velocity, latency and duration.16,17 

Ocular Structure

The ocular structure being tracked, and the type of light source used, will impact data acquisition.  Current eye trackers used in clinical practice can track the corneal reflection of light, the pupil, the iris, the corneal limbus (sclera/iris junction), the reflections of light from the lens and cornea (dual purkinje images) or the retina.11 When considering tracking, it is important to consider nuances across tracker type. For example, eye tracking based on corneal reflections is error prone due to ocular surface disease, like dry eye or corneal scars, which alter the reflection,18 and ambient room or variability in outdoor lighting can cause uneven illumination which can cause additional light spots impacting signal detection.19

Pupil/iris tracking has a range of nuances to consider because they are dynamic ocular structures and in constant flux. Pupil size impacts accuracy and precision20 and can vary based on sympathetic/parasympathetic tone, cognitive attention, drugs, emotional state, ambient room lighting, gaze position, and head position/head movement – all of which are factors in patients with TBI. Furthermore, pupil detection can vary due to iris color21 and ethnicity22, making data difficult to collect across diverse patient populations, particularly when data are collected with mobile phone cameras which use light in the visible spectrum – as opposed to infrared light sources.

Patients with TBI and migraine have abnormal pupillary dynamics26,27 and dry eye28 compared to normative control groups. It is unknown how accurate pupil/corneal reflection trackers are in a brain injury population as most device testing is done on an artificial eye or in a normal, healthy control population. Therefore, please note that confounding conditions such as dysautonomia, cranial nerve palsy, and nystagmus are not well normed in these devices.

Retinal trackers are the most precise and accurate of existing eye trackers.11 Benefits of the retina include uniformity in photoreceptor size across different demographics including age and ethnicity, and independence from influence by adrenergic/cholinergic medications, environmental testing conditions, mood, attention, and TBI. The tradeoff for precision in these devices is cost for the technology, reduced field of gaze, and necessity of a chin/forehead rest.

Visual Clarity

Finally, one factor that all eye trackers depend upon is the necessity of the patient to be able to see. Visual clarity is required for precise eye movements and reduced visual acuity can impact eye movements.32 This can be a problem in patients with reduced vision due to uncorrected refractive error, ocular disease, or accommodative dysfunction, the latter of which is present in up to 51% of patients post-concussion.30,33

Clinical Interpretation

Eye tracking as a biomarker is in its scientific infancy. At present, eye trackers are incapable of diagnosing the etiology of abnormal eye movements, and it still takes clinical judgment and experience to decipher if the oculomotor abnormality detected was secondary to the TBI, pre-existing or due to another neurologic disorder.32 While there is a great future potential for a role in artificial intelligence in pattern differentiation of eye movements and neurologic diagnosis, current software still relies on clinical interpretation.

Conclusion

  • Oculomotor assessment is a critical component to a comprehensive evaluation of patients with TBI and impacts patient quality of life and recovery.
  • The oculomotor system is an objective biomarker in all stages of brain injury and can be used for diagnosis, as well as recovery and monitoring of treatment interventions. 
  • Current technology has its disadvantages that clinicians should be aware of, but the benefits of eye tracking technology and the future of artificial intelligence will rapidly increase the utilization of this technology in clinical practice.

References

  1. Storey, E. P., Master, S. R., Lockyer, J. E., Podolak, O. E., Grady, M. F., & Master, C. L. (2017). Near Point of Convergence after Concussion in Children. Optometry and vision science : official publication of the American Academy of Optometry, 94(1), 96–100. https://doi.org/10.1097/OPX.0000000000000910
  2. Johnson, B., Hallett, M., & Slobounov, S. (2015). Follow-up evaluation of oculomotor performance with fMRI in the subacute phase of concussion. Neurology, 85(13), 1163–1166. https://doi.org/10.1212/WNL.0000000000001968
  3. Matuseviciene, G., Johansson, J., Möller, M., Godbolt, A. K., Pansell, T., & Deboussard, C. N. (2018). Longitudinal changes in oculomotor function in young adults with mild traumatic brain injury in Sweden: an exploratory prospective observational study. BMJ open, 8(2), e018734. https://doi.org/10.1136/bmjopen-2017-018734
  4. Capó-Aponte, J. E., Jorgensen-Wagers, K. L., Sosa, J. A., Walsh, D. V., Goodrich, G. L., Temme, L. A., & Riggs, D. W. (2017). Visual Dysfunctions at Different Stages after Blast and Non-blast Mild Traumatic Brain Injury. Optometry and vision science : official publication of the American Academy of Optometry, 94(1), 7–15. https://doi.org/10.1097/OPX.0000000000000825
  5. DuPrey, K. M., Webner, D., Lyons, A., Kucuk, C. H., Ellis, J. T., & Cronholm, P. F. (2017). Convergence Insufficiency Identifies Athletes at Risk of Prolonged Recovery From Sport-Related Concussion. The American journal of sports medicine, 45(10), 2388–2393. https://doi.org/10.1177/0363546517705640
  6. Cheever, K. M., McDevitt, J., Tierney, R., & Wright, W. G. (2018). Concussion Recovery Phase Affects Vestibular and Oculomotor Symptom Provocation. International journal of sports medicine, 39(2), 141–147. https://doi.org/10.1055/s-0043-118339
  7. Leonard, B. T., Kontos, A. P., Marchetti, G. F., Zhang, M., Eagle, S. R., Reecher, H. M., Bensinger, E. S., Snyder, V. C., Holland, C. L., Sheehy, C. K., & Rossi, E. A. (2021). Fixational eye movements following concussion. Journal of vision, 21(13), 11. https://doi.org/10.1167/jov.21.13.11
  8. Bin Zahid, A., Hubbard, M. E., Lockyer, J., Podolak, O., Dammavalam, V. M., Grady, M., Nance, M., Scheiman, M., Samadani, U., & Master, C. L. (2020). Eye Tracking as a Biomarker for Concussion in Children. Clinical journal of sport medicine : official journal of the Canadian Academy of Sport Medicine, 30(5), 433–443. https://doi.org/10.1097/JSM.0000000000000639
  9. Hunfalvay, M., Roberts, C. M., Murray, N., Tyagi, A., Kelly, H., & Bolte, T. (2019). Horizontal and vertical self-paced saccades as a diagnostic marker of traumatic brain injury. Concussion (London, England), 4(1), CNC60. https://doi.org/10.2217/cnc-2019-0001
  10. Nyström, M., Andersson, R., Holmqvist, K., & van de Weijer, J. (2013). The influence of calibration method and eye physiology on eyetracking data quality. Behavior research methods, 45(1), 272–288. https://doi.org/10.3758/s13428-012-0247-4
  11. Holmqvist, K., Örbom, S. L., Hooge, I., Niehorster, D. C., Alexander, R. G., Andersson, R., Benjamins, J. S., Blignaut, P., Brouwer, A. M., Chuang, L. L., Dalrymple, K. A., Drieghe, D., Dunn, M. J., Ettinger, U., Fiedler, S., Foulsham, T., van der Geest, J. N., Hansen, D. W., Hutton, S. B., Kasneci, E., … Hessels, R. S. (2022). Eye tracking: empirical foundations for a minimal reporting guideline. Behavior research methods, 10.3758/s13428-021-01762-8. Advance online publication. https://doi.org/10.3758/s13428-021-01762-8
  12. Cognolato, M., Atzori, M., & Müller, H. (2018). Head-mounted eye gaze tracking devices: An overview of modern devices and recent advances. Journal of rehabilitation and assistive technologies engineering, 5, 2055668318773991. https://doi.org/10.1177/2055668318773991
  13. Niehorster, D. C., Santini, T., Hessels, R. S., Hooge, I., Kasneci, E., & Nyström, M. (2020). The impact of slippage on the data quality of head-worn eye trackers. Behavior research methods, 52(3), 1140–1160. https://doi.org/10.3758/s13428-019-01307-0
  14. Carter, B. T., & Luke, S. G. (2020). Best practices in eye tracking research. International journal of psychophysiology : official journal of the International Organization of Psychophysiology, 155, 49–62. https://doi.org/10.1016/j.ijpsycho.2020.05.010
  15. Andersson, R., Nystrom, M., Holmqvist K, (2010). Sampling frequency and eye-tracking measures: how speed affects durations, latencies, and more. Journal of Eye Movement Research, 3(3), 6, 1-12.
  16. Inchingolo, P., & Spanio, M. (1985). On the identification and analysis of saccadic eye movements–a quantitative study of the processing procedures. IEEE transactions on bio-medical engineering, 32(9), 683–695. https://doi.org/10.1109/TBME.1985.325586
  17. Juhola, M., J¨antti, V., & Pyykk¨ o, I. (1985). Effect of sampling frequencies on computation of the maximum velocity of saccadic eye movements. Biological Cybernetics, 53(2), 67– 72.
  18. Goto, E., Dogru, M., Sato, E. A., Matsumoto, Y., Takano, Y., & Tsubota, K. (2011). The sparkle of the eye: the impact of ocular surface wetness on corneal light reflection. American journal of ophthalmology, 151(4), 691–696.e1. https://doi.org/10.1016/j.ajo.2010.09.027
  19. Wan Z., Xiong C., Chen W., & Zhang H, (2021). Robust and accurate pupil detection for head-mounted eye tracking. Computers and Electrical Engineering, 9, 1-14.
  20. Choe, K. W., Blake, R., & Lee, S. H. (2016). Pupil size dynamics during fixation impact the accuracy and precision of video-based gaze estimation. Vision research, 118, 48–59. https://doi.org/10.1016/j.visres.2014.12.018
  21. Hessels, R.S., Andersson, R., Hooge, I., Nystrom, M., Kemner., C. (2015). Consequences of eye color, positioning, and head movement for eye-tracking data quality in infant research. Infancy. 20(6), 601-633
  22. Blignaut, P., & Wium, D. (2014). Eye-tracking data quality as affected by ethnicity and experimental design. Behavior research methods, 46(1), 67–80. https://doi.org/10.3758/s13428-013-0343-0
  23. Hooge, I., Holmqvist, K., & Nyström, M. (2016). The pupil is faster than the corneal reflection (CR): Are video based pupil-CR eye trackers suitable for studying detailed dynamics of eye movements?. Vision research, 128, 6–18. https://doi.org/10.1016/j.visres.2016.09.002
  24. Nyström, M., Hooge, I., & Holmqvist, K. (2013). Post-saccadic oscillations in eye movement data recorded with pupil-based eye trackers reflect motion of the pupil inside the iris. Vision research, 92, 59–66. https://doi.org/10.1016/j.visres.2013.09.009
  25.  Nyström, M., Hooge, I., & Andersson, R. (2016). Pupil size influences the eye-tracker signal during saccades. Vision research, 121, 95–103. https://doi.org/10.1016/j.visres.2016.01.009
  26. Truong, J. Q., & Ciuffreda, K. J. (2016). Comparison of pupillary dynamics to light in the mild traumatic brain injury (mTBI) and normal populations. Brain injury, 30(11), 1378–1389. https://doi.org/10.1080/02699052.2016.1195922
  27. Harle, D. E., Wolffsohn, J. S., & Evans, B. J. (2005). The pupillary light reflex in migraine. Ophthalmic & physiological optics : the journal of the British College of Ophthalmic Opticians (Optometrists), 25(3), 240–245. https://doi.org/10.1111/j.1475-1313.2005.00291.x
  28. Lee C, Felix E, Levitt R, et al. Traumatic brain injury, dry eye and comorbid pain diagnoses in US veterans. Br J Ophthalmol 2018;102:667–673.
  29. Irving, E. L., González, E. G., Lillakas, L., Wareham, J., & McCarthy, T. (2011). Effect of stimulus type on the eye movements of children. Investigative ophthalmology & visual science, 52(2), 658–664. https://doi.org/10.1167/iovs.10-5480
  30. Merezhinskaya, N., Mallia, R. K., Park, D., Bryden, D. W., Mathur, K., & Barker, F. M., 2nd (2019). Visual Deficits and Dysfunctions Associated with Traumatic Brain Injury: A Systematic Review and Meta-analysis. Optometry and vision science : official publication of the American Academy of Optometry, 96(8), 542–555. https://doi.org/10.1097/OPX.0000000000001407
  31. Ghasia, F. F., Otero-Millan, J., & Shaikh, A. G. (2018). Abnormal fixational eye movements in strabismus. The British journal of ophthalmology, 102(2), 253–259. https://doi.org/10.1136/bjophthalmol-2017-310346
  32. Alexander, R. G., Macknik, S. L., & Martinez-Conde, S. (2018). Microsaccade Characteristics in Neurological and Ophthalmic Disease. Frontiers in neurology, 9, 144. https://doi.org/10.3389/fneur.2018.00144
  33. Master, C. L., Scheiman, M., Gallaway, M., Goodman, A., Robinson, R. L., Master, S. R., & Grady, M. F. (2016). Vision Diagnoses Are Common After Concussion in Adolescents. Clinical pediatrics, 55(3), 260–267. https://doi.org/10.1177/0009922815594367