A radiomics model developed from pre-metastatic brain magnetic resonance imaging to predict brain metastasis in non-small-cell lung cancer: Evidence to the seed-and-soil theory

2030 Background: Steven Paget’s “seed and soil” theory suggests that both the primary tumor and the target host organ play important roles in successful metastasis development. In this proof-of-concept study, we proposed that inter-individual differences exist in the brain’s congeniality for developing brain metastasis (BM), and validated the theory via developing a non-invasive radiomics BM prediction model. Methods: 256 non-small cell lung cancer (NSCLC) patients with no BM at baseline brain magnetic resonance imaging (MRI) were selected, 128 patients developed BM within three years after diagnosis and 128 remained BM-free. For radiomics analysis, both the BM and non-BM groups were randomly distributed into training and testing datasets at an 70%:30% ratio. Baseline brain MRI (representing the soil) and chest computed tomography (CT, representing the seed) radiomics features were extracted to develop the BM prediction models. We first developed radiomics models using the training dataset and subsequently validated the models in the testing dataset. A radiomics BM score (RadBM score) was generated and BM-free survival were compared between RadBM score-high and RadBM score-low groups. Results: The radiomics model developed from baseline brain MRI features alone can predict BM development in NSCLC patients. Furthermore, a fusion model integrating brain MRI features with primary tumor CT features (seed-and-soil model) provided synergetic effect and was more efficient in predicting BM (areas under the receiver operating characteristic curve 0.84[95% confidence interval: 0.80–0.89] and 0.80 [95% confidence interval: 0.71–0.88] in the training and testing datasets, respectively). BM-free survival was significantly shorter in the RadBM score-high group versus the RadBM score-low group (Log-rank P < 0.001). Cumulative BM rate at 3-year were 75.8% and 24.2% for RadBM score-high and RadBM score-low group, respectively. Conclusions: The results demonstrated that intrinsic features of a non-metastatic organ exert a significant impact on metastasis development, which is first-in-class in metastasis prediction studies and provided novel evidence for the "seed-and-soil" theory of tumor metastasis. Radiomics BM prediction model utilizing both pre-metastatic brain and primary tumor features might provide a useful tool for identifying NSCLC patients more prone to develop BM and providing individualized management for these patients.