Imaging modalities, including contrast-enhanced computed tomography (CE-CT) and 2-deoxy-2-[18F]fluoro-D-glucose positron emission tomography/computed tomography ([18F]FDG PET/CT), may be useful in the early prediction and monitoring of tyrosine kinase inhibitor (TKI) treatment effect in patients with gastrointestinal stromal tumor (GIST), according to a systematic review published in Diagnostics.
Studies showed CE-CT imaging features, including a tumor size of 5 cm or more, a lobulated tumor outline, mixed or exophytic growth patterns, and the presence of enlarged and tumor-feeding blood vessels, were predictive of poorer patient outcomes. High-risk GISTs also generally displayed heterogeneous enhancement patterns with areas of hypodensity potentially correlating with neovascularization or tumor necrosis, Ylva Weeda, of the Leiden University Medical Center in the Netherlands, and colleagues noted.
Fewer studies were available for [18F]FDG PET/CT imaging. However, some features of this imaging, including metabolic tumor volume and total lesion glycolysis, were correlated with more aggressive GIST tumors. [18F]FDG PET/CT was useful for observing reductions in maximum standardized uptake values within a week of TKI initiation, and these metabolic changes often preceded reductions in tumor size.
Read more about GIST prognosis
The use of radiomics and deep learning models with CE-CT imaging was found to provide high-performance predictive models for recurrence-free survival (RFS) and risk stratification. Several studies used texture radiomic features, including neighboring gray-tone difference matrix, gray level cooccurrence matrix, gray run-length matrix, and gray level size zonal matrix, to predict high-risk GISTs. Mutational status of KIT exon 11 was less predictable using imaging features, however.
“Radiomics is an emerging topic in medicine and shows potential for the prediction of RFS and risk stratifications in GISTs. However, future research should mainly focus on clinical utility, explainability, and correlation with actual tumor biology,” the authors said.
An initial literature search identified 599 articles, but through the study selection process, this number was reduced to 90 eligible articles for analysis. Of these 90 articles, roughly a quarter (n=23) were concerned with response monitoring, while the remaining studies focused on response prediction (n=67).
Weeda Y, Kalisvaart G, van Velden F, et al. Early prediction and monitoring of treatment response in gastrointestinal stromal tumors by means of imaging: a systematic review. Diagnostics. Published online November 7, 2022. doi:10.3390/diagnostics12112722