Abstract
BACKGROUND: Bone metastasis is a common complication in advanced malignancies, often resulting in severe pain and reduced quality of life. Radiopharmaceuticals like Strontium-89 ((89)Sr) are commonly used for palliative treatment to alleviate bone pain associated with metastases. This study explores the potential of radiomics analysis in predicting the effectiveness of (89)Sr treatment for pain relief in patients with bone metastases. METHODS: The study analyzed clinical and imaging data from 146 patients with bone metastases, specifically focusing on two types of lesions: osteolytic and osteoblastic. Pain relief was assessed by the step of the WHO pain ladder required for pain relief, along with a reduction in opioid dosage, indicating effective pain management. Based on exploratory analysis, a Bagging Decision Tree machine learning model was selected for outcome prediction in osteolytic lesions, while the XGBoost model was utilized for osteoblastic lesions. Both models leveraged radiomics features extracted from these lesions to improve predictive accuracy. Model performance was evaluated using the area under the receiver operating characteristic curve (AUC), along with sensitivity, specificity, accuracy, and calibration curves. RESULTS: The pain relief rate for osteolytic metastases was 58.33%, and for osteoblastic metastases, it was 62.16%. The Bagging Decision Tree model achieved an AUC of 0.991 in the training set and 0.889 in the test set for osteolytic lesions. For osteoblastic lesions, the XGBoost model yielded robust results, with an AUC of 0.970 in the training set and 0.958 in the test set. CONCLUSION: This study shows promise in predicting pain relief outcomes of (89)Sr treatment in patients with bone metastases.