# Can radiomics signatures and machine learning methods reinforce the revived role of 18F-NaF in metastatic bone disease?

## TL;DR

This study explores whether radiomic features from 18F-NaF PET/CT scans, combined with machine learning, can better distinguish true bone metastases from false positives.

## Contribution

The novel use of radiomics and machine learning to enhance diagnostic accuracy of 18F-NaF PET/CT in metastatic bone disease is demonstrated.

## Key findings

- XGBoost achieved the highest AUC of 0.86 in distinguishing true metastatic from false-positive lesions.
- Key radiomic features like SUVmax and SUVmean were identified as important predictors.
- Machine learning improved diagnostic accuracy compared to traditional methods.

## Abstract

To evaluate whether radiomic features extracted from 18F-NaF PET/CT scans, analyzed using machine learning (ML) methods, can improve the differentiation between true metastatic bone lesions (TP) and false-positive benign uptake (FP), thereby enhancing the diagnostic utility of 18F-NaF PET/CT.

This retrospective study included 62 patients with known primary malignancies who underwent 18F-NaF PET/CT. Lesions were classified as TP or FP based on consensus interpretation including follow-up. Patients were randomly split into training (n=41) and validation (n=21) groups. Radiomic features were extracted from PET images using LIFEx software. Feature selection (ANOVA, RFE) and ML model training (SVM, Random Forest, XGBoost) were performed. Model performance was evaluated using accuracy, specificity, sensitivity, and AUC, initially with a train/validation split and subsequently with 5-fold cross-validation incorporating feature engineering and hyperparameter tuning. Feature importance was assessed using SHAP.

Significant differences in SUVmax (p=0.006) and SUVmean (p=0.034) were observed between TP and FP lesions. Initial validation showed XGBoost performed best (AUC=0.78). After optimization and 5-fold cross-validation on the combined dataset (n=62), the tuned XGBoost model achieved the highest performance (Mean Accuracy: 85.7% ±2.9%, Mean AUC: 0.86), outperforming Random Forest (AUC: 0.79) and SVM (AUC: 0.74). SHAP analysis identified SUVmax, SUVmean, Voxel Volume Num, GLRLM RLNU, and Skew.

Radiomics-based machine learning classifiers, particularly XGBoost, demonstrated strong performance in distinguishing true metastatic from false-positive benign lesions on 18F-NaF PET/CT. Integrating radiomics and ML can potentially improve the diagnostic accuracy and robustness of 18F-NaF PET/CT for assessing bone metastases. Further validation in larger cohorts is warranted.