Radiomics Based on Brain-to-tumor Interface Enables Prediction of Metastatic Tumor Type of Brain Metastasis: a Proof-of-concept Study
Mingchen Jiang,Yiyao Sun,Chunna Yang,Zekun Wang,Ming Xie,Yan Wang,Dan Zhao,Yuqi Ding,Yan Zhang,Jie Liu,Huanhuan Chen,Xiran Jiang
DOI: https://doi.org/10.1007/s11547-024-01934-4
2024-01-01
La radiologia medica
Abstract:Early and accurate identification of the metastatic tumor types of brain metastasis (BM) is essential for appropriate treatment and management. A total of 450 patients were enrolled from two centers as a primary cohort who carry 764 BMs originated from non-small cell lung cancer (NSCLC, patient = 173, lesion = 187), small cell lung cancer (SCLC, patient = 84, lesion = 196), breast cancer (BC, patient = 119, lesion = 200), and gastrointestinal cancer (GIC, patient = 74, lesion = 181). A third center enrolled 28 patients who carry 67 BMs (NSCLC = 24, SCLC = 22, BC = 10, and GIC = 11) to form an external test cohort. All patients received contrast-enhanced T1-weighted (T1CE) and T2-weighted (T2W) MRI scans at 3.0 T before treatment. Radiomics features were calculated from BM and brain-to-tumor interface (BTI) region in the MRI image and screened using least absolute shrinkage and selection operator (LASSO) to construct the radiomics signature (RS). Volume of peritumor edema (VPE) was calculated and combined with RS to create a joint model. Performance of the models was assessed by receiver operating characteristic (ROC). The BTI-based RS showed better performance compared to BM-based RS. The combined models integrating BTI features and VPE can improve identification performance in AUCs in the training (LC/NLC vs. SCLC/NSCLC vs. BC/GIC, 0.803 vs. 0.949 vs. 0.918), internal validation (LC/NLC vs. SCLC/NSCLC vs. BC/GIC, 0.717 vs. 0.854 vs. 0.840), and external test (LC/NLC vs. SCLC/NSCLC vs. BC/GIC, 0.744 vs. 0.839 vs. 0.800) cohorts. This study indicated that BTI-based radiomics features and VPE are associated with the metastatic tumor types of BM.