Imaging tumor microenvironment heterogeneity in oral squamous cell carcinoma using IVIM DWI habitat analysis.

OBJECTIVE: The tumor microenvironment (TME) of oral squamous cell carcinoma (OSCC) is spatially heterogeneous and critically influences prognosis. Conventional magnetic resonance imaging (MRI) diffusion-weighted imaging (DWI) metrics average heterogeneity and obscure relevant biological intratumoral patterns. Habitat imaging enables voxel-wise characterization of diffusion-perfusion heterogeneity. MATERIALS AND METHODS: Eighty-four patients with pathologically confirmed OSCC were prospectively enrolled and underwent preoperative multi-b-value intravoxel incoherent motion (IVIM)-DWI. A voxel-wise habitat imaging framework was applied based on diffusion-related (Dt) and perfusion-related (f) parameters. Four habitats were defined using population-level Dt and f thresholds, corresponding to distinct combinations of diffusion and perfusion characteristics, to ensure standardized subregion definitions across patients. Tumor-stroma ratio (TSR), tumor-infiltrating lymphocytes (TILs), and cervical lymph node metastasis (CLNM) were assessed on histopathological sections. Habitat metrics, including subregional percentage, volume, and mean Dt and f values, were quantified and correlated with tumor-level pathological phenotypes. The predictive performance was evaluated. RESULTS: High-TSR tumors exhibited higher Dt value within hypocellular habitats (subregions 3 and 4), which independently predicted a stroma-rich phenotype. High-TIL tumors demonstrated lower Dt in a cellular-dominant, high-perfusion habitat (subregion 2) and a reduced percentage of hypocellular, low-perfusion habitat (subregion 3), consistent with immune-enriched microenvironments. CLNM-positive tumors showed overall habitat expansion, with increased volume of a putative stroma-dominant habitat independently associated with metastasis. Habitat-based models outperformed whole-tumor apparent diffusion coefficient and Dt for predicting TSR (area under the receiver operating characteristic curve (AUC) = 0.791) and TILs (AUC = 0.782). CONCLUSION: IVIM-DWI-based habitat imaging provides a noninvasive approach for characterizing diffusion-perfusion heterogeneity in OSCC and may complement histopathological assessment for preoperative risk stratification. KEY POINTS: Question: Can IVIM-DWI-based habitat imaging noninvasively characterize tumor microenvironment heterogeneity and identify imaging features associated with stromal composition, immune infiltration, and nodal metastasis in OSCC? FINDINGS: IVIM-DWI-based habitat imaging identified spatially distinct diffusion-perfusion habitats associated with TSR, TILs, and CLNM, with improved discrimination over conventional whole-tumor diffusion metrics. RELEVANCE STATEMENT: IVIM-DWI-based habitat imaging provides clinically relevant, noninvasive biomarkers of TME features, improving preoperative risk stratification and supporting more precise, individualized therapeutic decisionmaking in OSCC.