Radiomics model based on MRI to differentiate spinal multiple myeloma from metastases: A two-center study.

Area of Science:

• Radiology and Medical Imaging
• Oncology
• Machine Learning in Medicine

Background:

• Spinal multiple myeloma (MM) and metastases share similar imaging features, necessitating accurate differential diagnosis for effective treatment.
• Distinguishing between these conditions is crucial for guiding precision therapy in oncology.

Purpose of the Study:

• To develop and evaluate radiomics models for differentiating spinal multiple myeloma from spinal metastases using MRI.
• To compare the diagnostic performance of various machine learning models in this differentiation task.

Main Methods:

• A cohort of 263 patients (127 with spinal MM, 136 with spinal metastases) from two institutions was analyzed.
• Radiomics features were extracted from contrast-enhanced T1-weighted imaging (CET1) and T2-weighted imaging (T2WI) sequences.
• Machine learning models including Logistic Regression, AdaBoost, SVM, Random Forest, and MKL-SVM were trained and validated.

Main Results:

• The Random Forest model performed best on single sequences, with T2WI outperforming CET1.
• Integrating both CET1 and T2WI sequences significantly enhanced the diagnostic efficiency of all models.
• The multiple kernel learning based SVM (MKL-SVM) model achieved the highest performance, with an accuracy of 0.862, AUC of 0.870, and F1-score of 0.874.

Conclusions:

• Radiomics models derived from MRI demonstrate significant potential for differentiating spinal MM and metastases.
• These models offer promising prospects for improving individualized diagnosis and treatment strategies in clinical practice.