MENU

Click here to

×

Are you sure ?

Yes, do it No, cancel

The Impact of Multi-B-Value On Radiomic Features of MR Diffusion Weighted Imaging in Hepatocellular Carcinoma

J Zhang*, Q Qiu , J Duan , Y Yin , Shandong Cancer Hospital Affiliated to Shandong University, Jinan, Shandong

Presentations

(Wednesday, 8/1/2018) 10:30 AM - 11:00 AM

Room: Exhibit Hall | Forum 5

Purpose: To assess the impact of b-values on radiomic features of diffusion-weighted imaging and to determine the stable radiomic features between different b-values in hepatocellular carcinoma(HCC).

Methods: The diffusion-weighted imaging (DWI) with ten b-values (0, 20, 50, 100, 200, 400, 800, 1000, 1200 and 1500 s/mm², respectively) and T1WI images of 34 patients with HCC diagnosed by radiologists were performed on a 3.0T MR scanner (GE 3.0T Discovery). Then, we defined the volumes of interest (VOI) in b=0 s/mm² DWI combining T1WI sequences as reference, and then were mapped to other nine b-values DWI. For patients whose diameter of lesion ≥ 20×20 mm, three VOIs were defined in the region of cancer distributing in differencing surfaces and others drawn the whole contour. After that, 78 radiomic features were extracted by software of 3D Slicer Radiomics Extension. The robustness of radiomic features across different b-values was evaluated by percentage coefficient of variation (%COV) the features whose %COV < 30 were considered to be stable.

Results: Twenty radiomic features had low variation with varying b-values, including 2 IH features, 8 GLCM features, 6 GLRLM features and 4 GLSZM features. Meanwhile, the smallest variation feature pairs were extracted from DWI images with nearby b-values, b=0, 20 and 50 s/mm² image pairs.

Conclusion: Our study reveals that the variation in robustness of radiomic features extracted from DWI with different b-values. Texture features are more stable than intensity-based features and the features extracted from nearby b-value image will be more reproducible.

Keywords

MRI, Quantitative Imaging, Functional Imaging

Taxonomy

IM- MRI : Quantitative imaging/analysis

Contact Email