Classification of short and long term mild traumatic brain injury using computerized eye tracking

Area of Science:

• Neuroscience
• Ophthalmology
• Medical Technology

Background:

• Diagnosing brain injuries, particularly mild traumatic brain injury (mTBI), presents significant challenges.
• Objective and reliable assessment tools are crucial for understanding the impact of mTBI and persistent post-concussion syndrome (PPCS).

Purpose of the Study:

• To evaluate the usability and test-retest reliability of computerized eye-tracker assessments (CEAs).
• To assess the ability of CEAs to differentiate between healthy individuals, mTBI patients, and PPCS patients using machine learning.

Main Methods:

• CEAs assessed oculomotor function, visual attention, and selective attention, including tests like smooth-pursuit and the vestibulo-ocular reflex (VOR).
• Test-retest reliability was assessed in healthy adults, followed by machine learning model training using data from healthy, mTBI, and PPCS participants.

Main Results:

• CEAs demonstrated moderate to excellent reliability (ICC > .50 to .98) and satisfactory compliance.
• Machine learning models achieved reasonable balanced accuracy (Control: 0.83, mTBI: 0.66, PPCS: 0.76) with an AUC-ROC of 0.82.
• Key differentiating metrics included VOR (gaze stability), fixation vertical error, and pursuit measures.

Conclusions:

• Computerized eye-tracker assessments are reliable tools for evaluating patients with mTBI and PPCS.
• While promising, further development with larger datasets is needed to enhance diagnostic model accuracy for clinical application.