Author: Keyi Bian, Marco Caballo, Wenxiu Guo, Haijie Li, Jiao Li, Aidi Liu, Yue Ma, Ioannis Sechopoulos, Yafei Wang, Yaopan Wu, Zhaoxiang Ye, Yuwei Zhang, Yueqiang Zhu, Daan van den Oever 👨🔬
Affiliation: Radboud University Medical Center, Tianjin Medical University Cancer Institute & Hospital, Sun Yat-Sen University Cancer Center 🌍
Purpose: To develop and validate a nomogram integrating intra- and peritumoral radiomics of contrast-enhanced cone-beam breast CT (CE-CBBCT) and clinicopathologic features for predicting fluorescence in situ hybridization (FISH) status in HER2 2+ breast cancer.
Methods: This study included 199 patients with 252 HER2 2+ breast cancer lesions from two hospitals who underwent CE-CBBCT examination between 2017 and 2023. Patients from TCIH and SYSUCC comprised training (n=181) and test (n=71) cohorts, respectively. VOIs were delineated and radiomics features were extracted in intra- and peritumoral regions. RadScore derived from the two regions and their combination were calculated for building radiomics models, followed by interpretion using SHapley Additive exPlanations (SHAP) framework. A clinical model was constructed based on independent predictors. A nomogram was developed by integrating the best RadScore and independent clinicopathologic predictors. The predictive efficacy of these models was compared by ROC curve. The clinical utility of the nomogram was evaluated by decision curve analysis (DCA).
Results: The RadScore of combined intra- and peritumoral regions had the best AUC compared with that of each region alone. SHAP analysis revealed that three intratumoral and two peritumoral features contributed greatly to the predictive efficiency of the radiomics model. Lesion presence and ER status were selected as independent predictors for building the clinical model. In the test cohort, the nomogram achieved the best performance (nomogram/radiomics/clinical model, AUC=0.774/0.737/0.666, p<0.05). The DCA indicated that the nomogram had overall higher net benefit than the clinical model.
Conclusion: The nomogram could have potential in differentiating FISH-positive from -negative, and is helpful for accurately identifying FISH-positive in HER2 equivocal (2+) immunohistochemical result to distinguish between HER2-low and HER2-positive breast cancer.