Abstract
BACKGROUND AND AIM: Pyogenic liver abscess (PLA) is a devastating and potentially life-threatening disease globally, with Klebsiella pneumoniae liver abscess (KPLA) being the most prevalent in Asia. This study aims to develop an effective and comprehensive nomogram combining clinical and radiomics features for early prediction of KPLA. METHODS: 255 patients with PLA from 2013 to 2023 were enrolled and randomly divided into the training and validation cohorts at a 7:3 ratio. The differences between the two cohorts of patients were assessed via univariate analysis. The radiomics features were extracted from imaging data from enhanced CT of liver abscesses. The optimal radiomics features were filtered using the independent sample t-test and least absolute shrinkage and selection operator, and a radiomics score (Rad-score) was calculated by weighting their respective coefficients. Clinically independent predictors were identified from the clinical data and combined with the Rad-score to develop a nomogram by multivariate logistic regression. The predictive performance was evaluated using the area under the receiver operating characteristic curve (AUC), calibration curve, and clinical decision curve. RESULTS: The nomogram incorporated four clinical features of diabetes mellitus, cryptogenic liver abscess, C-reactive protein level, and splenomegaly, and the Rad-score that was constructed based on seven optimal radiomics features. It had an AUC of 0.929 (95% CI, 0.894-0.964) and 0.923 (95% CI, 0.864-0.981) in the training and validation cohorts, respectively. The calibration and decision curves showed that the nomogram had good agreement and clinical applicability. CONCLUSIONS: The clinical-radiomics nomogram performed well in predicting KPLA, hopefully serving as a reference for early diagnosis of KPLA.