MRI-based radiomic analysis for grading myxoid liposarcoma: a multisequence retrospective study

Area of Science:

• Radiology
• Oncology
• Medical Imaging

Background:

• Myxoid liposarcoma (MLS) grading is crucial for treatment and prognosis.
• Accurate histological grading of MLS can be challenging.

Purpose of the Study:

• To identify quantitative MRI features using radiomic analysis for predicting MLS histological grade.
• To develop predictive models for MLS grading.

Main Methods:

• Retrospective analysis of 57 MLS patients (30 low-grade, 27 high-grade).
• Extraction of 107 radiomic features from various MRI sequences (T1-WI, T2-WI, STIR, ADC, CE-T1-WI with/without FS).
• Development and cross-validation of predictive models, assessing performance via AUC.

Main Results:

• Radiomic analysis identified significant features across different MRI sequences.
• Predictive models based on T2-weighted imaging (T2-WI) and contrast-enhanced 3D (CE-3D) achieved the highest performance (AUC up to 0.88).
• Models showed reduced performance on external validation but improved with same-vendor data.

Conclusions:

• Pre-treatment MRI radiomics shows promise for predicting MLS histological grade.
• This approach offers quantitative, whole-tumor grading, complementing conventional methods.
• Further multicentric validation is needed for clinical application.