People in the United States have high incidence of mild traumatic brain injury (mTBI), especially the symptoms of concussion. However, there are questions about the traditional measurements of concussion. It is difficult to detect, and its key information may be easily ignored or become too subjective.This study measures the eye-tracking data by using VR and VR environments. We use the one-way univariate analyses of variance to examine the differences in performance on different ages, gender, and TBI severity between participants with concussion and controls. We not only check the effects of saccades, fixations, and reaction time but also make sure that three data can represent a biomarker for differentiating patients with a concussion or not.In addition, the analysis technology of electroencephalography (EEG) is used in this project, and the visual method in the VR environment is applied to generate non-invasive, visual stimulation to the experimenter to collect the response of the human brain. This technology is called Visual Evoked Potential (VEP) to analyze the brain waves of ordinary people generated by the stimulus, compare, and even combine artificial intelligence methods and eye movement values to detect concussion symptoms of mTBI. We hoped that the fusion technology of machine learning can combine the above two different analysis results to achieve the effect of optimizing the accuracy and convenience of concussion detection.