Machine Learning Model for Predicting Pheochromocytomas/Paragangliomas Surgery Difficulty: A Retrospective Cohort
Yubing Zhang1, Qikun Guo2, Shurong Li3
1Department of Urology, The First Affiliated Hospital, Sun Yat-sen University, Guangzhou, Guangdong, People's Republic of China.
Annals of Surgical Oncology
|May 9, 2025
Summary
A machine learning model using clinical and radiomic data accurately predicts pheochromocytoma and paraganglioma surgical difficulty. This aids in preoperative risk assessment and personalized surgical planning to minimize operative risks.
Area of Science:
- Endocrinology
- Oncology
- Medical Imaging
Background:
- Pheochromocytomas and paragangliomas (PPGLs) are rare neuroendocrine tumors.
- Surgical difficulty in PPGL resection can lead to increased complications.
- Accurate preoperative assessment of surgical difficulty is crucial for patient management.
Purpose of the Study:
- To develop and validate a machine learning (ML) model for predicting surgical difficulty in PPGLs.
- To integrate clinical and radiomic features for enhanced predictive accuracy.
- To compare the performance of ML models with varying feature sets.
Main Methods:
- Retrospective analysis of 212 patients with confirmed PPGLs.
- Development and comparison of seven ML models (including SVM, Random Forest) using clinical and radiomic data.
- Performance evaluation using AUC, accuracy, sensitivity, specificity, and F1 score; SHAP analysis for interpretability.
Main Results:
- The Support Vector Machine (SVM) model integrating clinical and radiomic features achieved the highest performance (AUC 0.96 training, 0.85 validation).
- The integrated SVM model significantly outperformed the clinical parameter-only model.
- Radiomic signature was the most influential predictor, followed by age, BMI, tumor diameter, and heart rate.
Conclusions:
- An SVM model combining clinical and radiomic features effectively predicts PPGL surgical difficulty.
- This model can aid in preoperative risk stratification and personalized surgical planning.
- The findings suggest a potential to reduce operative risks through improved preoperative assessment.
