What question did this study set out to answer?

This study aims to assess the effectiveness of NODDI habitat-based radiomics in distinguishing glioblastoma from solitary brain metastasis.

April 29, 2026Open Access

Cluster habitat-based diffusion MRI radiomics for differentiation of glioblastoma from solitary brain metastasis

Key Points

This study aims to assess the effectiveness of NODDI habitat-based radiomics in distinguishing glioblastoma from solitary brain metastasis.
Retrospective, two-center study involving 279 patients (196 glioblastoma, 83 solitary brain metastasis).
Performed k-means clustering on NODDI images and developed five different predictive models.
Evaluated performance using AUROC, calibration curves, and decision curve analysis.
The habitat model achieved a testing AUROC of 0.851, outperforming radiomics and clinical models.
The combined model with habitat features and clinical variables had the highest predictive performance (testing AUROC = 0.912).
Habitat-H2 subregion demonstrated strong discriminatory ability with a validation AUROC of 0.900.

Abstract

Abstract Objective Given the distinct intratumoral heterogeneity of glioblastoma (GB) and solitary brain metastasis (SBM), habitat-based radiomics derived from neurite orientation dispersion and density imaging (NODDI) may offer enhanced diagnostic value. This study aimed to evaluate NODDI habitat analysis in distinguishing GB from SBM. Materials and methods This retrospective, two-center study included 279 patients (196 GB, 83 SBM) who underwent 3-T magnetic resonance imaging (MRI), including T1-, T2-, and diffusion-weighted as well as contrast-enhanced T1-weighted sequences. K-means clustering was performed on NODDI images within the region of interest. Following feature extraction, five models were developed: habitat subregion, habitat, radiomics, clinical, and combined. Performance was evaluated using the area under the receiver operating characteristic curve (AUROC), calibration curves, and decision curve analysis. Results The region of interest was divided into three habitat subregions, with the Habitat-H2 subregion demonstrating strong discriminatory ability (validation AUROC = 0.900; testing AUROC = 0.828). Compared to the radiomics and clinical models, the habitat model containing the three subregions showed a higher discriminatory ability (validation AUROC = 0.929; testing AUROC = 0.851). The combined model, integrating habitat features and clinical variables (age) into a nomogram, achieved the highest predictive performance (validation AUROC = 0.931; testing AUROC = 0.912) and provided superior clinical value. Conclusion NODDI-based habitat MRI radiomics shows potential for differentiating GB from SBM, while integrating clinical variables is essential for optimal diagnostic performance. Relevance statement NODDI-based habitat radiomics aids differentiation between glioblastoma and solitary brain metastasis. Key Points Habitat analysis based on NODDI facilitates the differentiation between GB and SBM on MRI. The benefits of habitat analysis are maximized in the three clustered subregions. Habitat radiomics captures intratumoral heterogeneity that whole-tumor radiomics may miss. Graphical Abstract

AI에게 질문

Bookmark

View Full Paper