ScholarWorks@중앙대학교

초록

Objective Freezing of gait (FOG) significantly affects quality of life and increases the risk of falls in patients with Parkinson’s disease (PD). Although deep brain stimulation (DBS) of the globus pallidus interna (GPi) is effective in managing motor complications, its efficacy in treating FOG remains inconsistent. This study aimed to determine whether preoperative structural brain connectivity can predict both the presence of FOG and its postoperative improvement following GPi DBS. Methods We retrospectively analyzed 58 patients with PD who underwent GPi DBS. Preoperative diffusion tensor imaging was used to assess structural connectivity between the volume of activated tissue (VAT) and 82 cortical regions. Machine learning models were developed to predict baseline FOG and postoperative FOG improvement (defined as a ≥1- or ≥2-point reduction) using demographic and connectivity features. Results Machine learning models incorporating structural connectivity features between the VAT and cortical regions—including the prefrontal, cingulate, and premotor cortices—outperformed models based solely on demographic variables in predicting both the presence of preoperative FOG and postoperative improvement. For example, the support vector machine model to predict FOG improvement (≥1-point improvement) achieved an accuracy of 0.65 with demographic data alone, which increased to 0.77 with the addition of structural connectivity features. Similar performance enhancements were observed in sensitivity analyses using stricter FOG thresholds (≥2-point improvement). Conclusion Preoperative structural connectivity between the GPi and key cortical regions involved in cognitive control and motor planning predicts FOG responsiveness to DBS. These results highlight the utility of connectomic biomarkers for personalizing DBS strategies and optimizing therapeutic outcomes in patients with advanced PD.

키워드