Radiomics and Artificial Intelligence Applied to Carotid Ultrasound for Mortality Prediction in Sepsis: A Pilot Study

Background: Endothelial dysfunction plays a central role in the pathophysiology of sepsis but is rarely assessed in routine clinical practice.Carotid ultrasound is a widely available bedside tool; however, conventional analysis has limited sensitivity for detecting early sepsis alterations.This pilot study evaluated the feasibility of applying radiomics to carotid ultrasound images combined with Artificial Intelligence (AI) to predict mortality in patients with viral and bacterial sepsis.Methods: Clinical data and carotid ultrasound images from critically ill patients admitted to the intensive care unit were prospectively analyzed.Radiomic features were extracted from ultrasound images, and machine learning models were developed using clinical variables, radiomic features, and their combination.Model performance was assessed using area under the receiver operating characteristic curve (AUC), accuracy, sensitivity, precision, and F1-score.Results: A total of 37 patients were included (mean age 61.3 ± 14.6 years; 62.1% male), of whom 27 had viral sepsis due to COVID-19 and 10 had bacterial sepsis.Among the evaluated models, the Naive Bayes classifier showed the best performance when combining clinical and radiomic data, achieving an AUC of 0.97, accuracy of 94%, sensitivity of 98%, and precision of 94%.The inclusion of radiomic features increased model robustness by approximately 9% compared with the clinical model alone.Conclusion: Radiomics applied to carotid ultrasound imaging is feasible and demonstrates high predictive performance for mortality in patients with sepsis.The integration of radiomics and AI may enhance endothelial assessment and improve risk stratification in critically ill patients.