Multiparametric MRI-Based Radiomics for Identifying Primary Central Nervous System Diffuse Large B-cell Lymphomas'...

Area of Science:

• Neuroradiology
• Oncology
• Medical Imaging

Background:

• Primary central nervous system diffuse large B-cell lymphoma (PCNS-DLBCL) is an aggressive non-Hodgkin lymphoma.
• Accurate pathological subtyping (e.g., germinal center B-cell-like [GCB] vs. non-GCB) is crucial for treatment decisions.
• Distinguishing between PCNS-DLBCL subtypes preoperatively remains challenging.

Purpose of the Study:

• To evaluate the predictive capability of radiomics features from multiparametric MRI for identifying PCNS-DLBCL pathological subtypes.
• To compare the performance of radiomics models derived from different MRI sequences and combinations.

Main Methods:

• 186 PCNS-DLBCL patients (55 GCB, 131 non-GCB) were included.
• Radiomics features were extracted from segmented tumor regions across T1WI, T2WI, T2-FLAIR, CE-T1WI, and ADC maps.
• Radiomics models were built using individual sequences, combined sequences, and clinical features.

Main Results:

• Combined-sequence radiomics models generally outperformed single-sequence models.
• The apparent diffusion coefficient (ADC) and contrast-enhanced T1-weighted imaging (CE-T1WI) combined model achieved the highest discriminative power (AUC = 0.867).
• Incorporating more sequences or clinical features did not significantly improve model performance.

Conclusions:

• Multiparametric MRI-based radiomics holds significant value for predicting PCNS-DLBCL pathological subtypes.
• Radiomics analysis offers a non-invasive tool to aid in the preoperative subtyping of PCNS-DLBCL.