PET/CT Radiomics Integrated with Clinical Indexes as A Tool to Predict Ki67 in Breast Cancer: A Pilot Study

Abstract

Abstract Background: Breast cancer (BC) represents the second cause of cancer-related death among women globally, and Ki67 was demonstrated as an important predictive biomarker in worse survival and neoadjuvant treatment in BC. This study aims to investigate the value of radiomics features derived from 18F-FDG PET/CT combined with clinical characteristics in predicting Ki67 in patients with BC. Methods: A total of 114 patients diagnosed as BC and examined using 18F-FDG PET/CT were included in this study. Patients were randomly separated into a training set (n = 79, with 55 cases of Ki67 + and 24 cases of Ki67-) and a validation set (n = 35, with 24 cases of Ki67 + and 11 cases of Ki67-) at a ratio of 7:3. Thirteen clinical characteristics and 704 radiomics features were extracted, and the univariance logistic analysis, max-Relevance and Min-Redundancy, the least absolute shrinkage and selection operator regression, and the Spearman test were applied for feature selection. Three models were developed, including the clinical model, the radiomics model, and the combined model, and a nomogram of the combined model was constructed. The predictive performance of all three models was examined by the receiver operating characteristic (ROC) curve. Clinical utility was validated by decision curve analysis (DCA). Results: The N stage, tumor morphology, maximal standard uptake value, and the longest diameter were significantly different in Ki67 + and Ki67- groups (P < 0.05) and were selected as the most discriminative clinical features. Eight radiomics features were selected for the radiomics model. In total, 7 radiomics and the above 4 clinical characteristics were selected for the combined model. The AUC of the combined model in the training and test group was 0.90 (95% Confidence Interval (CI): 0.82–0.97) and 0.81 (95% CI: 0.64–0.99), respectively. The combined model significantly outperformed the radiomics model and the clinical model alone (P < 0.05). The DCA curve showed the advantages of the combined model over the clinical model and radiomics model. Conclusions: The radiomics-derived features combined with the clinical features could effectively predict Ki67 expression in BC based on PET/CT images. Trial registration: This study was registered at ClinicalTrials Gov (number NCT05826197) on 7th, May 2023.