04 / 2022 – 10 / 2022
Mild traumatic brain injuries (mTBI) or concussions have become a global healthcare crisis, with annual incidence rates reaching 600 per 100,000 individuals . A concussion is defined as a change in brain function that occurs after a force to the head, which may be accompanied by temporary loss of consciousness but is detected in awake individuals using neurologic and cognitive dysfunction measures . These measures are based on Computer Tomography (CT) scans in hospitals or on the patient’s subjective self-reported symptoms [1, 2].
Especially in sports when a CT scan is not feasible on the sideline the concussion assessment still relies on standardized but subjective assessment tools, hence, a timely diagnosis of concussion remains difficult [4, 5, 6]. Players who continue to play after suffering a head injury have a higher risk of developing persistent post-trauma syndrome (PTS) and life-threatening states, such as second impact syndrome and the possibility of later accelerated neurodegeneration [1, 8]. As a result, identifying and validating objective diagnostic methods that can assist doctors and physicians in evaluating head injuries is vital and urgent.
Current approaches can be divided into three main categories : cognitive, motor, and visualbased tools. The cognitive tools are commonly digital representations of the aforementioned subjective assessment tools. To increase accuracy the tools often rely on baseline measurements which makes deployment more complicated. The motor tools are based on close observation or on inertial measurement units to capture movements. However, it remains unclear how this data can be correlated to concussions. So further research has to be done. The visual tools rely on eye-tracking either by observation or by eye-tracking devices to analyze the saccadic eye movements. It was shown that when mTBI individuals did saccadic eye movement tasks, they had longer saccadic latencies than Control participants . However, there is still a lack of meaningful thresholds or measures that help to assess concussion solely on eye-tracking data . Nevertheless, it is well known that concussion can lead to focal neurological deficits in which the eyes cannot jointly fixate on a target . Because of that, a simple stereoscopic task can help to identify those deficits and has proven to be a good assessment tool for concussion . We want to combine the promising results of eye-tracking with a simple stereoscopic task to test for focal neurological deficits and thus broaden the visual-based assessments available so far.
The main goal of this thesis is to classify the presence of concussion based on altered visual behavior and a stereoscopic task. We intend to develop a portable off-the-shelf eye-tracked Virtual Reality(VR) concussion assessment tool to quickly and objectively screen for oculomotor dysfunctions (VR Oculomotor Test System, VR-OTS). On the screen, there will be four balls as stimuli appearing in sequence at nine gaze points: upper, lower, left, right, middle, upper left, lower left, upper right, and lower right. One of the four balls appears to be closer to the user which has to be identified by a button press. Data acquired during this task are the correctness of the task (whether the correct ball was identified) and the reaction time (time from stimulus presentation and button press). During the examination, eye behavior is tracked. We expect a difference in accuracy, reaction time, and eye behavior when comparing healthy controls and concussion patients. Comprehensive and reasonable data will be obtained through a large number of tests so that the data can be processed using machine learning (ML) and deep learning (DL) methods. Random Forest (RF) has been used in the past to investigate if such a model can accurately predict concussion based on eye behavior [10, 11, 12]. Hence, as a first step, we will also utilize RF but also other models like Support Vector Machine (SVM), K-Nearest-Neighbor (KNN) or Long-Short-Term Memory Networks (LSTM). The latter is especially interesting for eye-tracking data which are time-series data .
 D. Shukla. et al.: Mild traumatic brain injuries in adults.J. Neurosci. RuralPract. 1 (2) (2010) 82-88, https://doi.org/10.4103/0976-3147.71723. Publisher:Medknow Publications.
 Robert Graham. et al.: Sports-Related Concussions in Yough: Improving the Science, Changing the Culture. 2013.
 NCAA Sport Science Institute.: 2014-2015 NCAA Sports Medicine Handbook. National Collegiate Athletic Association, Indianapolis, IN..
 Paul McCrory. et al.: Consensus statement on concussion in sportsport-the 5th international conference on concussion in sport held in Berlin. October 2016. British Journal of Sports Medicine 51, 11 (2017),838-847. https://doi.org/10.1136/bjsports-2017-097699
 S. Herring. et al.: Selected issues in sport-related concussion (SRC|mild traumatic brain injury) for the team physician: a consensus statement. British journal of sports medicine 55 (22), pp. 1251â€“1261. DOI: 10.1136/bjsports-2021-104235.
 D. Powell. et al.: Sports related concussion: an emerging era in digital sports technology.npj Digit. Med. 4, 164 (2021). https://doi.org/10.1038/s41746-021-00538-w
 D. Kara, et al.: Detection of Mild Traumatic Brain Injury by a Virtual Reality System. Invest. Ophthalmol. Vis. Sci. 2019;60(9):2304.
 T.D. Stein. et al.: Concussion in chronic traumatic encephalopathy. Curr. Pain Headache Rep. 19 (10) (2015).
 B.P. Johnson. et al.:A closer look at visually guided saccades in autism and Asperger’s disorder. Front. Integr. Neurosci. 6 (2012).
 Y. Mao, et al.: Disease Classification Based on Eye Movement Features With Decision Tree and Random Forest. Frontiers in Neuroscience 14, p. 798. DOI: 10.3389/fnins.2020.00798.
 K. Tirdad, et al.: Machine learning-based approach to analyze saccadic eye movement in patients with mild traumatic brain injury. Computer Methods and Programs in Biomedicine Update 1, p. 100026, 2021. DOI: 10.1016/j.cmpbup.2021.100026.
 H. Wu, Hsiaokuang, et al.: Portable VR-based Concussion Diagnostics of Mild Traumatic Brain Injury. 2021 IEEE 3rd Eurasia Conference on IOT, Communication and Engineering (ECICE), Yunlin, Taiwan, 10/29/2021 – 10/31/2021: IEEE, pp. 21â€“25.
 F. Karim, et al.: LSTM Fully Convolutional Networks for Time Series Classification. IEEE Access 6, pp. 1662â€“1669, 2018. DOI: 10.1109/ACCESS.2017.2779939.