الملخص
Objective:To investigate the value of radiomics signatures based on 18F-FDG PET/CT for predicting molecular classification and Ki-67 expression of breast cancer. Methods:A total of 134 female patients ((55.4±13.3) years) who underwent 18F-FDG PET/CT examination and were diagnosed with breast cancer by pathology in the First Affiliated Hospital of Soochow University from April 2016 to May 2023 were retrospectively enrolled. LIFEx software was used to extract radiomics features and the least absolute shrinkage and selection operator (LASSO) algorithm and independent-sample t test were used to screen potentially meaningful features and calculate the radiomics score, which were considered as radiomics models. Clinical characteristics were selected by supervised logistic regression and clinical models were established. Radiomics features and clinical characteristics were incorporated to logistic regression analysis to establish combined models. ROC curves were drawn and the differences among AUCs were analyzed by Delong test. Results:Among 134 patients, 22 were with triple negative breast cancer (TNBC), 47 were human epidermal growth factor receptor 2 (HER2) over-expression type, 37 were Luminal A type and the rest 28 were Luminal B type. The expression of Ki-67 was high in 85 patients, and was low in the rest 49 patients. The AUCs (95% CI) of the combined models for predicting TNBC, HER2 overexpression type, Luminal A type and Ki-67 expression were 0.843(0.770-0.900), 0.808(0.723-0.876), 0.825(0.711-0.908) and 0.836(0.762-0.894), respectively, which were higher than those of clinical models ( z values: 1.97-3.06, all P<0.05). Conclusion:The predictive model combining radiomics signatures based on 18F-FDG PET/CT and clinical characteristics can well predict the molecular classification and Ki-67 expression level of breast cancer.