A nomogram-based model and ultrasonic radiomic features for gallbladder polyp classification

Background and aim: Gallbladder polyps (GBPs) are relatively common. Many studies have attempted to distinguish between benign and neoplastic GBPs to identify early-stage gallbladder carcinoma. We have established an accurate neoplastic predictive model and evaluated the effectiveness of radiomics in predicting malignancy in patients with GBPs.

Methods: A total of 503 patients confirmed through postoperative pathology were included in this retrospective study. Clinical information and ultrasonographic findings were retrospectively analyzed. The model was constructed from independent risk factors using Spearman correlation and logistic regression analysis of a training cohort of 250 GBP patients, and its efficacy was verified using an internal validation group of 253 consecutive patients through the receiver operating characteristic curve (ROC). The area of GBPs was delimited manually, and the texture features of ultrasound images were analyzed using correlation and ROC analysis.

Results: Independent predictors, including age, gallstones, carcinoembryonic antigen, polyp size, and sessile shape, were incorporated into the nomogram model for the neoplastic potential of GBPs. Compared with other proposed prediction methods, the established nomogram model showed good discrimination ability in the training group (area under the curve [AUC]: 0.865) and validation group (AUC: 0.845). Regarding ultrasonic radiomics, the minimum caliper diameter was identified as the only independent predictor (AUC: 0.841).

Conclusions: Our preoperative nomogram model can successfully evaluate the neoplastic potential of GBPs using simple clinical data, and our study verified the use of radiomics in GBP identification, which may be valuable for avoiding unnecessary surgery in patients.