Predicting progression-free survival in sarcoma using MRI-based automatic segmentation models and radiomics...

Area of Science:

• Oncology
• Radiology
• Medical Imaging

Background:

• Sarcomas are aggressive cancers with high recurrence rates.
• Accurate prediction of progression-free survival (PFS) is crucial for treatment planning.

Purpose of the Study:

• To develop and validate a radiomics signature for estimating sarcoma PFS.
• To assess the performance of a radiomics-based nomogram in predicting PFS.

Main Methods:

• Retrospective analysis of 202 sarcoma patients using pre-treatment MRI.
• Development of a radiomics signature using Least Absolute Shrinkage and Selection Operator (LASSO) Cox regression.
• Construction of a nomogram combining radiomics and clinical features.

Main Results:

• The ROI-Net framework achieved high segmentation accuracy (Dice coefficient 0.820).
• A 21-feature radiomics signature effectively distinguished high-risk patients.
• The Radiomics-T1WI-Clinical model demonstrated strong predictive performance (AUC up to 0.947).

Conclusions:

• The ROI-Net framework provides reliable image segmentation.
• Radiomics features and the combined nomogram show promise for predicting sarcoma PFS.
• This approach can potentially aid in clinical decision-making for sarcoma patients.