To develop and validate a multimodal machine learning model to predict postoperative worsening of freezing of gait questionnaire (FOGQ) scores in patients with Parkinson’s disease (PD) undergoing subthalamic nucleus deep brain stimulation (STN-DBS), emphasizing the validation aspect.
Key Findings:
The LightGBM model achieved the highest AUC of 0.917 for predicting FOGQ deterioration, indicating strong predictive performance.
Multimodal data integration significantly enhanced predictive accuracy.
Interpretation:
The multimodal LightGBM model effectively discriminates between patients with and without postoperative FOGQ deterioration, suggesting its utility in preoperative risk stratification and treatment planning, with potential clinical implications.
Limitations:
The study is retrospective and may be subject to selection bias.
External validation in prospective multicenter cohorts is necessary.
The sample size may limit the generalizability of the findings.
Conclusion:
The findings underscore the importance of integrating clinical, structural, and radiomic features for predicting postoperative outcomes in PD patients undergoing STN-DBS, highlighting the need for external validation.
by Min Xu, Shuhong Mei, Shuming Huang, Longyuan Gu, Yuting Zhang, Siyan Chen, Yuyao Tian, Li Du, Hui Zhao, Zixuan Zhang, Ruyi Chen, Guiyun Cui, Wei Zhang, Jie Zu
