JMIR Publications

ABSTRACT

Background:

Quantitative pupillometry has been used in mTBI with changes in pupil reactivity noted after blast injury, chronic mTBI and sports-related concussion.

Objective:

We evaluated the diagnostic capabilities of a smartphone-based digitial pupillometer to differentiate patients in the emergency room with mTBI from controls.

Methods:

Adult patients diagnosed with acute mTBI with normal neuroimaging were evaluated in an emergency department within 36 hours of injury. Healthy adults without mTBI were enrolled as controls. The PupilScreen smartphone pupillometer was used to measure the pupillary light reflex (PLR), and quantitative curve morphological parameters of the PLR were compared between mTBI and healthy controls. To address the class imbalance present in our sample, a synthetic minority oversampling technique (SMOTE) was applied. All possible combinations of PLR parameters produced by the smartphone pupillometer were then applied as features to four binary classification machine learning algorithms: Random Forest, k-nearest neighbors, support vector machine, and logistic regression. A 10-fold cross validation technique stratified by cohort was used to produce accuracy, sensitivity, specificity, area under the curve (AUC), and F1 score metrics for the classification of mTBI versus healthy subjects.

Results:

Acute mTBI patients (n=12) were 33% female, mean age 54.1 years, and 58% Caucasian with median Glasgow Coma Scale (GCS) of 15. Healthy patients (n=132) were 67% female, mean age 36 years, 64% Caucasian and median GCS of 15. Significant differences were observed in PLR recordings between healthy controls and acute mTBI patients in the following PLR parameters: Percent change (34±8.3 vs 26±7.9, p<0.007), minimum pupillary diameter (34.8±6.1 vs 29.7±6.1, p<0.007), maximum pupillary diameter (53.6±12.4 vs 40.9±11.9, p<0.007), and mean constriction velocity (11.5±5.0 vs 6.8±3.0, p<0.007) between cohorts. After SMOTE, both cohorts had a sample size of 132 recordings. The best performing binary classification model was a random forest model using latency, percent change, maximum diameter, minimum diameter, mean constriction velocity, and maximum constriction velocity PLR parameters as features. This model produced an overall accuracy of 93.5%, sensitivity of 96.2%, specificity of 90.9%, AUC of 0.936, and F1 score of 93.7% for differentiating between pupillary changes in mTBI and healthy subjects.

Conclusions:

Quantitative smartphone pupillometry may be a useful tool in the diagnosis of acute mTBI.