Artificial Intelligence of Neuropsychological Tests for the Prediction and Verification of Decline in Gait Parameters in Patients with Mild Cognitive Impairment

Background: Mild cognitive impairment (MCI) is considered a transitional state between normal aging and very early dementia. Increasing evidence reveals gait and cognition are inter-related in older adults with MCI. Therefore, it is important to find reliable biomarkers for these MCI patients, which can be utilized as an indicator for early detection and intervention. Methods: The deterioration of cognitive function will affect the patient's walking ability; thus, we conduct a two-stage study with comprehensive neuropsychological testing and a portable device for gait analysis at the beginning and repeated gait analysis six months later to evaluate gait deterioration. By machine learning using neuropsychological testing scores as the input feature parameters, a classification model capable of predicting the gait performance of MCI patients can be obtained. Results: Machine learning is capable of predicting several gait features of the MCI patients, such as reduction in walking speed (with up to 81.82% accuracy), increase in the time of the timed up and go (TUG) test (with up to 66.67% accuracy), and reduction in vertical jump height (with up to 69.23% accuracy) based on the predictive neuropsychological testing scores. Conclusion: Overall, the neuropsychological testing is predictive of gait decline, especially of walking speed, followed by vertical jump height in MCI patients. Therefore, the highest correlation among gait parameters in MCI patients could be the walking speed.