Computed tomography texture-based radiomics analysis in gallbladder cancer: initial experience

Abstract

Aim of the study: To investigate computed tomography (CT) texture parameters in suspected gallbladder cancer (GBC) and assess its utility in predicting histopathological grade and overall survival.

Material and methods: This retrospective pilot study included consecutive patients with clinically suspected GBC. CT images, clinical, and histological or cytological data were retrieved from the database. CT images were reviewed by two radiologists. A single axial CT section in the portal venous phase was selected for texture analysis. Radiomic feature extraction was done using commercially available research software.

Results: Thirty-eight patients (31 females, mean age 53.1 years) were included. Malignancy was confirmed in 29 patients in histopathology or cytology analysis, and the rest had no features of malignancy. Exophytic gallbladder mass with associated gallbladder wall thickening was present in 22 (58%) patients. Lymph nodal, liver, and omental metastases were present in 10, 1, and 3 patients, respectively. The mean overall survival was 9.7 months. There were significant differences in mean and kurtosis at medium texture scales to differentiate moderately differentiated and poorly differentiated adenocarcinoma (p < 0.05). The only texture parameter that was significantly associated with survival was kurtosis (p = 0.020) at medium texture scales. In multivariate analysis, factors found to be significantly associated with length of overall survival were mean number of positive pixels (p = 0.02), skewness (p = -0.046), kurtosis (0.018), and standard deviation (p = 0.045).

Conclusions: Our preliminary results highlight the potential utility of CT texture-based radiomics analysis in patients with GBC. Medium texture scale parameters including both mean and kurtosis, or kurtosis alone, may help predict the histological grade and survival, respectively.

Keywords: CT texture analysis; cancer; gallbladder; histology; radiomics.

Copyright © 2021 Clinical and Experimental Hepatology.

Figures

Fig. 1

Contrast enhanced axial CT image of a female patient with pathologically proven gallbladder cancer with texture features. A) CT image showing region of interest drawn around tumour (blue line) and corresponding images of fine, medium, and coarse textures. B) Histogram derived from image showing pixel distribution at filter value of 2 mm. C) Table of texture parameters across the different SSF values: 0 – conventional, 2 mm – fine texture scale, 3-5 mm – medium texture scales, and 6 mm – coarse texture scale

Fig. 2

Flowchart of patient recruitment

Fig. 3

Morphological types of gallbladder cancer. Axial contrast CT images depict asymmetric gallbladder wall thickening at fundus with associated exophytic mass in segment IV of liver (A, white arrow), heterogenous intraluminal polypoidal lesion ~ 2.5 cm in maximum dimension (B, white arrow) and heterogenous attenuation mass in gallbladder fossa with gallbladder not separately visualized (C, white arrow). D shows heterogenous exophytic mass with multiple calcific foci within (white arrow). All the above lesions were pathologically proven adenocarcinoma with good differentiation (B), moderate differentiation (C) and not otherwise specified (A, D), respectively, histological grades. Note is made of a periportal lymph node (arrowhead), with a CBD stent anterior to it (A)

Fig. 4

Boxplots showing the trends in the mean value at SSF3 and 4 (A, B) and kurtosis at SSF4 and 5 (C, D)

Fig. 5

Kaplan-Meier curves showing significant difference in overall survival for (A) kurtosis at SSF = 4 and (B) kurtosis at SSF = 5