Radiomics and artificial intelligence for soft-tissue sarcomas: Current status and perspectives

Area of Science:

• Oncologic Imaging
• Radiology
• Medical Artificial Intelligence

Background:

• Soft tissue sarcomas (STS) are rare, heterogeneous mesenchymal malignancies requiring advanced diagnostic tools.
• Current radiological assessment relies on imaging but can be limited in predicting STS behavior.
• Radiomics and AI offer novel approaches to extract quantitative imaging features for improved analysis.

Purpose of the Study:

• To review the current research status of radiomics and AI in the radiological assessment of STS.
• To detail the development and application of radiomics and deep learning in STS since 2010.
• To discuss the potential of these technologies in predicting STS characteristics and patient outcomes.

Main Methods:

• Explanation of radiomics principles, including image post-processing and feature extraction using machine learning.
• Review of deep learning algorithms, particularly convolutional neural networks, applied to STS.
• Focus on CT and MRI data for radiomics analysis in STS.

Main Results:

• Radiomics and deep radiomics models successfully discriminate between benign tumors and STS.
• Predictive models for histologic grade, treatment response, survival, and metastasis probability have been developed.
• Studies demonstrate the potential for radiomics in assessing STS prognosis and treatment efficacy.

Conclusions:

• Radiomics has shown significant potential in retrospective studies for STS assessment.
• Standardization, open-science initiatives, and independent databases are key for clinical validation.
• Future research should integrate radiomics with other '-omics' data for a comprehensive understanding of STS.