CT-based radiomics nomogram for differentiating dedifferentiated liposarcoma from well-differentiated liposarcoma

Area of Science:

• Radiology
• Oncology
• Medical Imaging

Background:

• Retroperitoneal liposarcoma (RLPS) classification is crucial for treatment.
• Accurate preoperative differentiation of RLPS subtypes remains challenging.

Purpose of the Study:

• To develop a predictive model for RLPS pathological classification using radiomics features from plain CT scans.
• To enhance preoperative planning and guide tailored treatment strategies for RLPS patients.

Main Methods:

• A retrospective study of 114 RLPS patients.
• Extraction and selection of radiomics features using LASSO regression.
• Development and validation of a nomogram for predicting dedifferentiated liposarcoma (DDLPS).

Main Results:

• Higher Ki-67 and unclear tumor boundaries identified as predictors for DDLPS.
• A five-feature radiomics nomogram achieved AUCs of 0.91 (training) and 0.89 (validation).
• The nomogram outperformed radiologist evaluation and provided greater clinical benefit.

Conclusions:

• The developed radiomics nomogram significantly improves preoperative differentiation of RLPS subtypes.
• This tool aids in precise diagnosis and personalized management of Retroperitoneal liposarcoma.