Assessing Quantitative Radiomic Features in DLBCL

A study found no discernable difference in the discrimination capability of radiomic features, which may predict diffuse large B-cell lymphoma (DLBCL) outcomes, and different segmentation methods, which are used to calculate metabolic tumor volume (MTV). The results appeared in the Journal of Nuclear Medicine. 

In this study, researchers matched 50 baseline 18F-fluorodeoxyglucose positron emission tomography computed tomography (PET/CT) scans of DLBCL patients who experienced disease progression or relapse within two years after diagnosis. These scans were then matched on uptake time and reconstruction method with 50 baseline PET/CT scans of DLBCL patients without disease progression. The scans were assessed 6 semi-automatic segmentation methods (standardized uptake value (SUV)4.0, SUV2.5, 41% of the maximum SUV, 50% of the SUVpeak, majority vote (MV)2 and MV3, respectively). These segmentation methods yielded 490 radiomics features at patient level and 486 features for the largest lesion. The research team used the intra-class correlation (ICC) agreement to calculated for each method compared to SUV4.0.

According to the results, the percentage of features yielding an ICC ≥0.75 compared to the SUV4.0 segmentation was lowest for A50P both at patient level and for the largest lesion, with 77.3% and 66.7% of the features yielding an ICC ≥0.75. The researchers noted that features were not highly correlated with MTV, with at least 435 features at patient level and 409 features for the largest lesion for all segmentation methods with a correlation coefficient <0.7.

“Even though there are differences in the actual radiomics feature values derived and selected features between segmentation methods, there is no substantial difference in the discriminative power of radiomics features between segmentation methods,” the researchers concluded.