18F-FDG PET/MRI-based liver and bone marrow organomics combined with systemic inflammatory response for predicting...

To evaluate the role of liver and bone marrow organomics features derived from staging 18F-fluorodeoxyglucose (18F-FDG) PET/MRI in predicting pathological complete response (pCR) in breast cancer patients who underwent surgery after neoadjuvant therapy.

In this retrospective study, breast cancer patients who underwent pretreatment staging with 18F-FDG PET/MRI, received neoadjuvant therapy, and subsequently underwent surgery were included. Clinicopathological characteristics and systemic inflammatory markers were recorded. Organomics features were extracted from the liver and L4 vertebra bone marrow regions on 18F-FDG PET images using LIFEx software. Least Absolute Shrinkage and Selection Operator regression was applied for radiomics feature selection, and a radiomics score (Radscore) was generated. Logistic regression analyses were performed to identify predictors of pCR. The diagnostic performances of the Radscore, Clinical, and Clinical-Radscore Combined models were evaluated using receiver operating characteristic analysis.

A total of 87 breast cancer patients were included, and pCR was observed in 33 patients (37.9%). In multivariable analysis, progesterone receptor status, human epidermal growth factor receptor 2 status, primary tumor size, and Radscore were identified as independent predictors of pCR. The Radscore yielded an area under the curve (AUC) of 0.722 [95% confidence interval (CI) = 0.625-0.819], whereas the Clinical and Clinical-Radscore Combined models achieved AUCs of 0.850 (95% CI = 0.765-0.934) and 0.904 (95% CI = 0.843-0.965), respectively. The Clinical-Radscore Combined Model demonstrated higher diagnostic performance than both the other models.

Liver and bone marrow organomics features derived from staging 18F-FDG PET/MRI may provide valuable information for predicting pCR in breast cancer patients undergoing neoadjuvant therapy.