Scholarlink

Purpose <p>While systematic reviews have summarized qualitative risk factors for postoperative cerebellar mutism syndrome (pCMS), a critical gap remains in translating these findings into a practical, quantitative tool for individualized preoperative prediction. This study aims to bridge this gap by developing and validating the first preoperative predictive model based on novel quantitative imaging indices to accurately estimate the risk of pCMS.</p> Methods <p>We retrospectively analyzed data from 73 pediatric patients who underwent posterior fossa surgery. We defined and measured a set of innovative quantitative imaging indices (including coronal index CORd, sagittal index SAGa, axial indices OAXa, OAXd, and a composite index OAXd*CORd) from preoperative MRIs. Anatomical invasion sites were also evaluated. Multivariate logistic regression was used to integrate these features into a predictive model. The model’s performance was compared with two previously published tools (Rotterdam scale and Zhang model) using discrimination metrics, calibration curves, and decision curve analysis.</p> Results <p>Invasion of the middle cerebellar peduncle (<i>p</i> = 0.001), superior cerebellar peduncle (<i>p</i> &lt; 0.001), and brainstem (<i>p</i> = 0.003) were significant anatomical risk factors. The quantitative indices SAGa, CORd, OAXa, OAXd, and OAXd*CORd showed significant differences between pCMS and non-pCMS groups (<i>p</i> &lt; 0.05). Our model had the highest AUC (0.850), sensitivity (92.31%) and negative predictive value (97.62%). Rotterdam scale had the highest overall accuracy (83.56%) and specificity (86.67%) of all three models. Our model yielded the highest net benefit at lower thresholds of risk (10–25%).</p> Conclusion <p>This study moves beyond existing qualitative risk factor lists by establishing a highly accurate, quantitative predictive model for pCMS. Comparison with existing tools demonstrates that our model offers superior sensitivity for identifying high-risk patients. The novel imaging indices provide refined insights into the pathophysiological mechanisms, potentially involving cerebello-thalamic pathways. This model serves as a practical clinical tool for preoperative risk assessment, surgical planning, and patient counseling.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Development and validation of a preoperative MRI-based predictive model for cerebellar mutism syndrome in pediatric posterior fossa tumors

  • 期刊论文

  • 发表日期：

    2026年4月23日

查看全文
  • Pei-Jun Xie,
  • Pei-Gang Lin,
  • Ru-Yu Ai,
  • Min-Ting Ye,
  • Qing-Jun Hu,
  • Shao-Qun Li,
  • Ming-Yao Lai,
  • Lin-Bo Cai,
  • Juan Li

摘要

Purpose

While systematic reviews have summarized qualitative risk factors for postoperative cerebellar mutism syndrome (pCMS), a critical gap remains in translating these findings into a practical, quantitative tool for individualized preoperative prediction. This study aims to bridge this gap by developing and validating the first preoperative predictive model based on novel quantitative imaging indices to accurately estimate the risk of pCMS.

Methods

We retrospectively analyzed data from 73 pediatric patients who underwent posterior fossa surgery. We defined and measured a set of innovative quantitative imaging indices (including coronal index CORd, sagittal index SAGa, axial indices OAXa, OAXd, and a composite index OAXd*CORd) from preoperative MRIs. Anatomical invasion sites were also evaluated. Multivariate logistic regression was used to integrate these features into a predictive model. The model’s performance was compared with two previously published tools (Rotterdam scale and Zhang model) using discrimination metrics, calibration curves, and decision curve analysis.

Results

Invasion of the middle cerebellar peduncle (p = 0.001), superior cerebellar peduncle (p < 0.001), and brainstem (p = 0.003) were significant anatomical risk factors. The quantitative indices SAGa, CORd, OAXa, OAXd, and OAXd*CORd showed significant differences between pCMS and non-pCMS groups (p < 0.05). Our model had the highest AUC (0.850), sensitivity (92.31%) and negative predictive value (97.62%). Rotterdam scale had the highest overall accuracy (83.56%) and specificity (86.67%) of all three models. Our model yielded the highest net benefit at lower thresholds of risk (10–25%).

Conclusion

This study moves beyond existing qualitative risk factor lists by establishing a highly accurate, quantitative predictive model for pCMS. Comparison with existing tools demonstrates that our model offers superior sensitivity for identifying high-risk patients. The novel imaging indices provide refined insights into the pathophysiological mechanisms, potentially involving cerebello-thalamic pathways. This model serves as a practical clinical tool for preoperative risk assessment, surgical planning, and patient counseling.