Integrating CT radiomics and morphologic features for preoperative risk stratification of gastrointestinal stromal tumours

A study conducted by researchers from Mansoura University aimed to evaluate the potential of merging CT morphological and radiomics characteristics to estimate the risk of gastrointestinal stromal tumours (GISTs) in patients. The research included 92 patients who had been confirmed to have GISTs through pathology. During the study, 42 radiomics features were extracted from the tumours observed in the portal venous phase of the CT scans.

The researchers conducted univariate analyses, comparing each morphological and radiomics characteristic between patients classified as low-risk and moderate/high-risk. After analyzing the data, three multivariate regression models were developed, utilizing either morphological features, radiomics features, or a combination of both to determine the most predictive variables and the optimal model for predicting GIST malignancy risk.

The study found that the presence of tumour necrosis and tumour vessels, when using CT morphologic features, were significant indicators for distinguishing between low-risk and moderate/high-risk GISTs, with an area under the curve (AUC) of 0.81. In a model utilizing CT radiomics features, skewness, total energy, and GLCM_Idmn (Gray-level co-occurrence matrix – inverse difference moment normalized) were significant indicators for differentiating between the risk groups, resulting in an AUC of 0.89.

When both sets of features were combined in a single model, the presence of tumour vessels and GLCM_Idmn proved significant in distinguishing between the risk groups, yielding an AUC of 0.90.

The researchers concluded that the integration of CT morphologic and radiomics features provided a valuable method for differentiating low-risk from moderate/high-risk GISTs, demonstrating high diagnostic performance. This combined approach has the potential to guide personalized treatment strategies and determine patient eligibility for adjuvant imatinib therapy.

The study is the first of its kind to integrate both CT morphological and radiomics features into a unified predictive model for risk stratification of GISTs. The findings were published under a Creative Commons Attribution 4.0 International License, allowing for free use, sharing, adaptation, distribution, and reproduction in any medium or format, as long as proper attribution to the original authors and source is given.