CT-based radiomics nomogram for predicting leukopenia in surgically treated stage IB-IIA cervical cancer following (chemo) radiotherapy

Moderate-to-severe leukopenia is a significant adverse effect in stage IB-IIA cervical cancer (CC) patients receiving postoperative radiotherapy/chemoradiotherapy (PORT/CRT). We aimed to develop and internally validate a CT-based radiomics nomogram that incorporates features of the psoas muscle at the third lumbar vertebra to predict this risk. This single-center, retrospective exploratory study included 150 patients randomly divided into training ( n = 106) and internal validation ( n = 44) cohorts using a single random split. Eighteen clinical variables and 2,153 radiomic features extracted from planning CT scans were analyzed. Key radiomic features were selected using least absolute shrinkage and selection operator (LASSO) binary logistic regression and incorporated into a radiomics score (rad-score) model. Multivariable analysis identified relevant clinical variables (clinical model), which, along with radiomic signatures, were used to construct the nomogram. Model performance, including that of the rad-score, clinical, and combined nomogram models, was assessed using receiver operating characteristic (ROC) curves, area under the ROC curve (AUC) values, and additional evaluation metrics. The LASSO regression identified 10 radiomic features, forming the basis of the rad-score model. Multivariable analysis indicated that the Nutritional Risk Screening 2002 tool and receipt of chemotherapy were independent clinical risk factors. The AUCs of the rad-score, clinical, and nomogram models were 0.743, 0.753, and 0.830, respectively, in the training cohort. In the internal validation cohort, the corresponding AUCs were 0.664 (95% CI, 0.491–0.818), 0.586 (95% CI, 0.401–0.756), and 0.800 (95% CI, 0.641–0.908), respectively. The nomogram showed potential for predicting moderate-to-severe leukopenia in patients with CC undergoing PORT/CRT. These findings are hypothesis-generating and based on internal validation only. Independent multicenter validation is required before routine clinical use.