Tumor heterogeneity in gastrointestinal stromal tumors of the small bowel: volumetric CT texture analysis as a potential biomarker for risk stratification

Background: To explore whether volumetric CT texture analysis (CTTA) can serve as a potential imaging biomarker for risk stratification of small bowel gastrointestinal stromal tumors (small bowel-GISTs). Methods: A total of 90 patients with small bowel-GISTs were retrospectively reviewed, of these, 26 were rated as high risk, 13 as intermediate risk, and 51 as low or very low risk. Histogram parameters extracted from CT images were compared among small bowel-GISTs with different risk levels by using one-way analysis of variance. Receiver operating characteristics (ROCs) and areas under the curve (AUCs) were analyzed to determine optimal histogram parameters for stratifying tumor risk. Results: Significant differences in mean attenuation, 10th, 25th, 50th, 75th and 90th percentile attenuation, and entropy were found among high, intermediate, and low risk small bowel-GISTs (p <= 0.001). Mean attenuation, 10th, 25th, 50th, 75th and 90th percentile attenuation, and entropy derived from arterial phase and venous phase images correlated significantly with risk levels (r = 0.403-0.594, r = 0.386-0.593, respectively). Entropy in venous phase reached the highest accuracy (AUC = 0.830, p < 0.001) for differentiating low risk from intermediate to high risk small bowel-GISTs, with a cut-off value of 5.98, and the corresponding sensitivity and specificity were 82.4 and 74.4%, respectively. Conclusions: Volumetric CT texture features, especially entropy, may potentially serve as biomarkers for risk stratification of small bowel-GISTs.