Radiomic features from MRI distinguish myxomas from myxofibrosarcomas

Area of Science:

• Oncology
• Radiology
• Pathology

Background:

• Myxoid tumors present diagnostic challenges, particularly distinguishing benign myxomas from malignant myxofibrosarcomas.
• Histological heterogeneity in myxofibrosarcomas can lead to misdiagnosis on core needle biopsies, mimicking myxomas.
• Current differentiation methods like FISH and immunohistochemistry are not always sufficient for myxoma versus myxofibrosarcoma differentiation.

Purpose of the Study:

• To evaluate the efficacy of T1-weighted signal intensity (T1SI), tumor volume, and radiomic features from MRI in differentiating myxomas from myxofibrosarcomas.
• To compare the diagnostic performance of these MRI-derived parameters.

Main Methods:

• Analysis of MRI scans from 56 patients (29 myxomas, 27 myxofibrosarcomas).
• Extraction of 89 radiomic features, alongside T1SI and tumor volume.
• Development and validation of random forests classifiers using leave-one-out cross-validation.

Main Results:

• Myxomas exhibited lower normalized T1SI compared to myxofibrosarcomas (p=0.006).
• The radiomic features classifier achieved an AUC of 0.885 (accuracy 0.839, sensitivity 0.852, specificity 0.828).
• Radiomic features significantly outperformed T1SI (AUC 0.713) and tumor volume (AUC 0.838) in classification accuracy.

Conclusions:

• Myxofibrosarcomas are generally larger and exhibit higher T1SI and distinct shape characteristics compared to myxomas.
• Radiomic features derived from MRI provide superior performance in differentiating myxomas from myxofibrosarcomas.
• This advanced imaging analysis offers a promising non-invasive tool for improved diagnostic accuracy of these rare tumors.