Multiparametric‐MRI‐Based Radiomics Model for Differentiating Primary Central Nervous System Lymphoma From Glioblastoma: Development and Cross‐Vendor Validation

Abstract

Background: Preoperative differentiation of primary central nervous system lymphoma (PCNSL) from glioblastoma (GBM) is important to guide neurosurgical decision-making. Purpose: To validate the generalization ability of radiomics models based on multiparametric-MRI (MP-MRI) for differentiating PCNSL from GBM. Study Type: Retrospective. Population: In all, 240 patients with GBM (n = 129) or PCNSL (n = 111). Field Strength/Sequence: 3.0T scanners (two vendors). Sequences: fluid-attenuation inversion recovery, diffusionweighted imaging (DWI), and contrast-enhanced T 1-weighted imaging (CE-T 1 WI). Apparent diffusion coefficients (ADCs) were derived from DWI. Assessment: Cross-vendor and mixed-vendor validation were conducted. In cross-vendor validation, the training set was 149 patients' data from vendor 1, and test set was 91 patients' data from vendor 2. In mixed-vendor validation, a training set was 80% of data from both vendors, and the test set remained at 20% of data. Single and multisequence radiomics models were built. The diagnoses by radiologists with 5, 10, and 20 years' experience were obtained. The integrated models were built combining the diagnoses by the best-performing radiomics model and each radiologist. Model performance was validated in the test set using area under the ROC curve (AUC). Histological results were used as the reference standard. Statistical Tests: DeLong test: differences between AUCs. U-test: differences of numerical variables. Fisher's exact test: differences of categorical variables. Results: In cross-vendor and mixed-vendor validation, the combination of CE-T 1 WI and ADC produced the bestperforming radiomics model, with AUC of 0.943 vs. 0.935, P = 0.854. The integrated models had higher AUCs than radiologists, with 5 (0.975 vs. 0.891, P = 0.002 and 0.995 vs. 0.885, P = 0.007), 10 (0.975 vs. 0.913, P = 0.029 and 0.995 vs. 0.900, P = 0.030), and 20 (0.975 vs. 0.945, P = 0.179 and 0.995 vs. 0.923, P = 0.046) years' experiences. Data Conclusion: Radiomics for differentiating PCNSL from GBM was generalizable. The model combining MP-MRI and radiologists' diagnoses had superior performance compared to the radiologists alone. Level of Evidence: 4 Technical Efficacy Stage: 2

Radiomic Based Machine Learning Performance for a Three Class Problem in Neuro-Oncology: Time to Test the Waters?

“…In addition, if upsampling is performed (for example from 0.5 × 0.5 × 5 mm 3 to 0.5 × 0.5 × 0.5 mm 3 ), there is a risk of introducing artificial information by inferencing a large number of voxels between slices.” [ 18 ]. As such, we performed standardized anisotropic resampling for all MRI sequences to ensure reproducibility as also performed in prior MRI radiomic studies [ 19 , 20 ]. Moreover, radiomic features have also been shown to be robust to different levels of pixel spacing and interpolation [ 21 ].…”

“…[18]. As such, we performed standardized anisotropic resampling for all MRI sequences to ensure reproducibility as also performed in prior MRI radiomic studies [19,20]. Moreover, radiomic features have also been shown to be robust to different levels of pixel spacing and interpolation [21].…”

Deep Learning for Automatic Differential Diagnosis of Primary Central Nervous System Lymphoma and Glioblastoma: Multi‐Parametric Magnetic Resonance Imaging Based Convolutional Neural Network Model

“…Radiomics is a subset of AI techniques that employs both statistical analysis and texture analysis to extract a large number of quantitative image features from the delineated tumor region of interest (ROI), followed by using a classifier to build diagnostic models 12 . Radiomics has shown great potential in building models that are capable of differentiating PCNSL from GBM 10,13–19 . However, radiomics requires precise tumor ROI delineation, which is time‐consuming, and there are delineation variabilities between observers.…”

“…12 Radiomics has shown great potential in building models that are capable of differentiating PCNSL from GBM. 10,[13][14][15][16][17][18][19] However, radiomics requires precise tumor ROI delineation, which is time-consuming, and there are delineation variabilities between observers. This drawback hinders the application of radiomics in clinical practice.…”

Progress of MRI Radiomics in Hepatocellular Carcinoma

“…However, the evaluation of imaging features may be subjective because radiologists have different experiences and different familiarities with the system(44,45). Radiomics has important application value in the diagnosis of solid tumors because it uses advanced image processing technology to extract high-throughput data and quantitative analysis of tumor behavior and heterogeneity(6,(46)(47)(48)(49)(50)(51).Radiomics signatures based on conventional precontrast T1-weighted imaging, postcontrast T1-weighted imaging, T2-weighted imaging, diffusion-weighted imaging (DWI), and intravoxel incoherent motion (IVIM), whether alone or in combination with clinical data, are all valuable for HCC differentiation (52-59), and their differentiation efficiency is almost equal to that of experienced radiologists (10-year experience)(52). HCC, intrahepatic cholangiocarcinoma (ICC), and HCC-ICC have common risk factors(60,61), and their typical qualitative MRI features may overlap(24,(62)(63)(64).…”