Predizione della percentuale di cellule ad anello con castone nei Ca gastrici di istotipo diffuso: ruolo della radiomica applicata alle immagini TC.

BACKGROUND In 2019, a Consensus updated the histopathological classification of gastric neoplasms, particularly refining the characterization of Diffuse-type neoplasms by Lauren. Prognostic studies revealed that a low percentage of signet ring cells (SRC <10%) is linked to a fourfold increase in cancer-specific mortality at five years, contrary to their historical perception as a negative prognostic factor. Since SRCs are heterogeneously distributed within tumors, their assessment via biopsy alone is insufficient. Radiomics, a quantitative imaging technique, offers the potential to non-invasively predict SRC percentages by extracting advanced imaging features, aiding pre-treatment decision-making. AIM This study aims to evaluate gastric neoplasms using computed tomography (CT) imaging to predict the percentage of signet ring cells during the staging process, thereby providing a non-invasive biomarker for risk stratification. MATERIALS AND METHODS The cohort included patients who underwent surgical resection at the Oncologic Surgery Department of Siena University Hospital. All patients were treatment-naïve (no neoadjuvant therapy), and biological samples were available from an approved tissue biobank. Histopathological reclassification was performed in collaboration with pathologists, adhering to the 2019 WHO classification criteria and the guidelines of the European Chapter of the IGCA to assess SRC percentages. Morphological analysis was conducted based on established radiological features and independently reviewed by two radiologists with differing levels of expertise. Radiomic feature extraction and segmentation were performed using open-source platforms, including 3D-Slicer and pyradiomics. Image standardization algorithms were applied to enhance the robustness of feature analysis. All statistical analyses, including model development, were conducted using SPSS and RStudio. RESULTS A total of 44 patients were included in the analysis. Significant prognostic differences emerged based on SRC percentages, with populations exhibiting <10% SRC demonstrating a markedly poorer prognosis (p=0.029). The morphological analysis yielded a predictive model with an AUC of 0.799, indicating moderate diagnostic performance. Radiomic analysis revealed the potential for overfitting; however, the application of LASSO regression enabled the development of a robust model with an AUC of 0.834. The integration of morphological and radiomic features further improved predictive performance, resulting in a combined model with an AUC of 0.838. DISCUSSION The findings underscore the potential of radiomics to enhance predictive accuracy for assessing SRC percentages in gastric neoplasms, outperforming traditional morphological analysis methods. Moreover, the integration of radiomic and morphological parameters demonstrated incremental benefits, highlighting the value of a multimodal approach. Future research should explore the integration of radiomics with advanced imaging technologies, such as dual-energy CT, which could further refine predictive models. However, the lack of dual-energy CT imaging in many patients within this cohort represents a limitation of the present study. Addressing this gap could pave the way for more precise, non-invasive prognostic tools in the management of gastric cancer.