Researchers have discovered that quantitative computed tomography (CT) texture analysis is a promising and noninvasive method for the classification of gastric tumors and prediction of gastric adenocarcinoma differentiation, according to a study published in the Japanese Journal of Radiology. 

Gastric tumors, such as gastrointestinal stromal tumors (GISTs), are routinely diagnosed using endoscopic biopsy, an invasive procedure. However, that method is uncomfortable for the patient and cannot be used to assess extra-serosal involvement.

CT scans, on the other hand, are commonly used preoperatively to stage gastric lesions. They are suitable for assessing neighboring structures of the tumor and any involvement of lymph nodes. CT texture analysis is a method of assessing pixel distribution and intensities in digital CT images, allowing users to objectively compare the characteristics of tumors that are not readily distinguishable to the naked eye.


Continue Reading

Read more about GIST diagnosis

The purpose of this study was to evaluate the effectiveness of CT texture analysis for distinguishing between gastric adenocarcinomas, lymphomas, and GISTs, as well as its effectiveness in distinguishing between the differentiation grades for gastric adenocarcinomas. Researchers recruited patients who had a preliminary diagnosis of gastric carcinomas from their archives between January 2014 and June 2019. All the patients included in the study had undergone pretreatment CT scans with intravenous and oral contrast. 

The contrast-enhanced venous phase CT images of 163 patients were retrospectively analyzed. Among them, 125 had gastric adenocarcinoma, 12 had gastric lymphoma, and 26 had GISTs. To minimize the impact of different CT scanners possibly yielding different results, researchers performed gray-level discretization, gray-level normalization, and pixel-size adjustment of the images. Researchers determined the region of interest from the axial slice that represented the largest cross-sectional lesion area. Forty texture parameters were also selected from each patient. 

Researchers were then able to compare the texture features between the different tumor subtypes, as well as the different grades of gastric adenocarcinoma. Texture features that were statistically significant were combined into one single parameter by logistic regression analysis. Receiver-operating characteristic curve (ROC) analysis was then used to differentiate tumor subtypes by measuring the specificity and sensitivity of these features and the combined parameter.

The findings of this study show that the classification of different gastric carcinomas and the prediction of different grades of adenocarcinomas using CT texture analysis yielded highly successful results with high specificity and sensitivity. The researchers of this study concluded that CT texture analysis is a noninvasive and promising tool that warrants larger patient studies in the future. 

Reference

Zeydanli T, Kilic HK. Performance of quantitative CT texture analysis in differentiation of gastric tumors. Jpn J Radiol. Published online July 25, 2021. doi:10.1007/s11604-021-01181-x