JoVE - Journal of Visualized Experiments
JoVE Visualize
Contact Us

Performance of Radiomics and Deep Learning Models in Predicting Distant Metastases in Soft Tissue Sarcomas: A Systematic Review and Meta-analysis

  • 0Department of Radiology, Division of Musculoskeletal Imaging and Intervention, University of Washington, Seattle, WA (P.M., S.H., H.S.K., A.A., M.C.).
Academic Radiology +

|

July 12, 2025

|

July 12, 2025

View abstract on PubMed

Summary

This summary is machine-generated.

Radiomics and deep learning models show promise for predicting soft tissue sarcoma (STS) metastasis. Deep learning approaches outperform traditional methods, aiding in identifying high-risk patients for targeted monitoring.

Area Of Science

  • Medical Imaging
  • Oncology
  • Artificial Intelligence

Background

  • Predicting distant metastases in soft tissue sarcomas (STS) is crucial for patient management.
  • Radiomics and deep learning (DL) models offer potential for metastasis prediction, but their accuracy needs evaluation.

Purpose Of The Study

  • To assess the diagnostic performance of radiomics and DL models in predicting STS metastases.
  • To analyze pooled sensitivity and specificity of these predictive models.

Main Methods

  • A meta-analysis following PRISMA guidelines, searching PubMed, Web of Science, and Embase.
  • Random-effects model to estimate pooled AUC, sensitivity, and specificity.
  • Subgroup analyses based on imaging modality, feature extraction (DL vs. handcrafted), and clinical features.

Main Results

  • Nineteen studies (1712 patients) included; pooled AUC for metastasis prediction was 0.88.
  • DL-based radiomics models demonstrated significantly higher sensitivity than handcrafted radiomics models (p < 0.01).
  • No significant performance difference was observed when including clinical features (AUC: 0.90 vs. 0.88, p = 0.99).

Conclusions

  • Radiomics models, particularly DL approaches, show significant potential for predicting STS metastasis.
  • These models can assist clinicians in identifying high-risk patients for targeted monitoring.
  • Further research is needed to address heterogeneity, external validation, and standardize protocols for clinical integration.

Keywords: