Impact of kinematics and kinetics on classification of dual-task gait

Early detection of Alzheimer’s Disease and Related Disorders (ADRD) has been a focus of research with the hope that early intervention may improve clinical outcomes. The manifestation of motor impairment in early stages of ADRD has led to the inclusion of gait assessments (typically qualitative or focused only on spatiotemporal gait features) in clinical evaluations for these diseases. This study aims to develop the groundwork for a classification tool to improve early detection of ADRD using biomechanical gait features by determining if machine learning algorithms can decode different levels of cognitive load in healthy individuals. A dual-task paradigm was used to simulate cognitive impairment in 40 healthy adults, with single-task walking trials representing normal, healthy gait. The Paced Auditory Serial Addition Task was administered at two different inter-stimulus intervals of 2.4 s and 1.6 s to manipulate cognitive load (Dual₁ and Dual₂ conditions, respectively). The primary hypothesis of this study is that using kinematic and kinetic gait features will improve classification performance compared to spatiotemporal gait features during single-task and dual-task gait classification among healthy adults. Repeated Measures ANOVA showed significant changes in 13 different gait features across all three levels of cognitive load (Single, Dual₁, and Dual₂). Three supervised machine learning algorithms (Partial Least Square Discriminant Analysis, Linear Discriminant Analysis, and Neural Network Pattern Recognition) were used to classify data points using a series of different gait feature sets, and performance was based on the area under the curve (AUC) of the receiver operating characteristic curve, which offers a combined measure of sensitivity and specificity on a 0-1 scale with a chance level of 0.5. Machine learning classification yielded AUC up to 0.865 for the Single vs Dual classification task (identifying presence of cognitive load), and up to 0.761 for classification across all conditions (identifying level of cognitive load). The results here show the ability to classify gait based on cognitive load with above-chance sensitivity and specificity using gait parameters and machine learning classifiers.