Metastasis risk prediction model in osteosarcoma using metabolic imaging phenotypes: A multivariable radiomics model

Area of Science:

• Oncology
• Radiology
• Medical Imaging

Background:

• Osteosarcoma (OS) is the most common primary bone cancer with high heterogeneity.
• Recurrence rates after initial treatment are 30-40%, with poor long-term survival (20%) for recurrent cases.
• Early metastasis prediction is critical for optimizing OS management and improving survival.

Purpose of the Study:

• To develop and evaluate a radiomics model for predicting metastasis risk in osteosarcoma patients.
• To utilize metabolic imaging phenotypes from 18F-FDG-PET scans for risk stratification.
• To identify imaging features that can guide treatment decisions in OS.

Main Methods:

• Eighty-three osteosarcoma patients undergoing surgery and chemotherapy were included.
• Pretreatment 18F-FDG-PET scans were analyzed, extracting 45 radiomic features.
• A multivariable logistic regression model was developed using cross-validation, selecting SUVmax and GLZLM-SZLGE features.

Main Results:

• The logistic regression model (Z = -1.23 + 1.53*SUVmax + 1.68*GLZLM-SZLGE) showed significant p-values for both SUVmax (0.0462) and GLZLM-SZLGE (0.0154).
• The final model achieved an Area Under the Curve (AUC) of 0.80.
• Cross-validation demonstrated a sensitivity of 0.66 and a specificity of 0.88 for metastasis prediction.

Conclusions:

• SUVmax and GLZLM-SZLGE from 18F-FDG-PET scans are independent predictors of metastasis risk in osteosarcoma.
• These metabolic imaging phenotypes correlate with metastatic potential, offering insights into disease progression.
• The developed radiomics model can enhance patient outcomes by identifying high-risk individuals for more aggressive treatment strategies.