Geometry Matters: Quantifying 3D-Fractality and Lacunarity in Tumor Subcomponents Distinguishes Between Brain Metastases and Gliomas

Motivation: The radiographic appearance of brain-metastases substantially overlaps with that of gliomas, 'the primary-tumors of the brain'. Currently, there is no definitive quantitative-MRI-index to distinguish between brain-metastases (arising-from-breast and -lung cancers) and gliomas. Goal(s): To develop an MRI-based-quantitative-platform that discriminates between brain-metastases and low and high-grade-gliomas using non-euclidean-geometric-measures of tumor-subcomponents. Approach: 3D-Fractality and Lacunarity, as non-euclidean-geometric-measures were quantified for the enhancing, non-enhancing, and edema subcomponents from preoperative-MRI of brain-metastases and low and high-grade-gliomas. Results: A lower 3D-Fractality of the non-enhancing and edema subcomponents in brain-metastases, compared to gliomas, is unique to brain-metastases. The combination of these two-tumor-subcomponents provided highly-accurate discrimination between brain-metastases and gliomas. Impact: This study introduces fractal-based tumor-geometry-metrics as innovative, non-invasive imaging-signature distinguishing brain-metastases (arising-from-breast and -lung cancers) and gliomas. Integrating 3D-fractality and lacunarity measurements of tumor-subcomponents in machine-learning-models yielded high-accuracy, precise-differentiation between brain-metastases and gliomas, thus reducing biopsy-dependency, enhancing noninvasive-differential-diagnosis, and prognostication.