Abstract
STUDY DESIGN: This is a cross-sectional study. OBJECTIVE: To examine the differences in gait patterns between cervical spondylotic myelopathy (CSM) patients and normal controls using visual recognition-based gait analysis and plantar pressure techniques, and to further analyze the differences in gait between different levels of disease severity as well as between gender groups. The development and progression of CSM symptoms are gradual and obscure. Although previous studies have objectively assessed CSM-specific gait patterns using motion cameras as well as mechanical platforms, these methods have limitations such as limited metrics that can be analyzed or inconvenience for simple screening. Therefore, there is a need to develop effective screening methods. METHODS: A total of 154 asymptomatic volunteers and 147 patients with CSM were recruited. Gait parameters were obtained by capturing video of the patient while walking via a smartphone-based computer vision algorithm (using the OpenPose framework), and plantar pressure distributions were measured using wireless sensor-equipped insoles (sampling rate 50-100 Hz). Gender, age, and body mass index were employed as matching variables, resulting in 102 matched pairs. An independent-samples t-test was used to analyze between-group differences, and analysis of variance was utilized for comparisons among three or more groups. Receiver operating characteristic (ROC) curve analysis, area under the curve (AUC), and the Youden index were employed to determine cut-off values for gait-related indicators. RESULTS: At a self-selected step speed, the CSM group exhibited significantly slower step speed and stride speed, shorter step length and stride length, a prolonged gait period, and an increased proportion of stance and double-support phases alongside a reduced swing phase. Multiple kinematic and kinetic parameters-such as hip flexion angle and ankle dorsiflexion angle-were markedly reduced. Medial plantar pressure was higher, whereas lateral plantar pressure was lower (P < 0.05). The stance phase demonstrated the highest diagnostic accuracy (AUC = 0.944, sensitivity = 88.9%, specificity = 89.2%), suggesting its potential as a principal screening variable. The optimal threshold for the stance phase was 64.75 (%). CONCLUSION: CSM patients differed significantly from normal subjects in terms of gait speed, stride speed, step length, stride length, gait period, hip flexion angle, ankle dorsiflexion angle, and medial and lateral plantar pressure. The percentage of stance phase in the entire gait period was greater than 64.75%, which can be used as the main gait parameter to assist in the diagnosis of CSM.