Background <p>To determine the feasibility of diffusion-relaxation correlation spectroscopic imaging in identifying tumoral differentiation profile and predicting cervical lymph node metastasis (CLNM) in oral tongue squamous cell carcinoma (OTSCC).</p> Materials and methods <p>This prospective study enrolled fifty-seven OTSCC patients who underwent preoperative head and neck magnetic resonance imaging (MRI). Scans with multi <i>b</i>-values (0–1500 s/mm²) and multi-echo times (7–150 ms) were performed to generate normalized diffusion-T2 spectra. Tumor maximal diameter and depth of invasion were measured. Tumors were segmented into five compartments (<i>V</i>_A to <i>V</i>_E) with metrics compared across normal controls, CLNM-, and CLNM+ groups. Pathological parameters such as tumor-stroma ratio (TSR), perineural invasion, Ki-67, tumor p53 protein, and cyclin-dependent kinase inhibitor p16 were evaluated. Correlations between MRI metrics and pathological parameters were assessed. Predictors of CLNM+ were identified using logistic regression analysis, and the predictive performance was evaluated using receiver operating characteristic analysis.</p> Results <p>Thirty-four patients were assigned to the CLNM+ group and 23 to the CLNM- group. CLNM+ patients showed larger tumor maximal diameters, deeper invasion, increased <i>V</i>_B and <i>V</i>_D, and decreased <i>V</i>_A compared to CLNM- patients. <i>V</i>_B exhibited strong positive correlations with perineural invasion and depth of invasion, while <i>V</i>_D correlated positively with TSR. Moreover, <i>V</i>_B and depth of invasion were independent prognostic factors for CLNM+, and their combined model achieved the highest predictive performance.</p> Conclusion <p>Diffusion-relaxation correlation spectroscopic imaging marked a significant advancement in the diagnostic and prognostic assessment of OTSCC, offering detailed tumor characterization and improving CLNM+ prediction, with great potential for accurate and non-invasive evaluation.</p> Relevance statement <p>Diffusion-relaxation correlation spectroscopic imaging metrics (<i>V</i>_B and <i>V</i>_D) characterized tumor heterogeneity and correlated with pathological biomarkers, making it a promising non-invasive tool for enhancing preoperative decisions and reducing unnecessary lymph node dissections in clinical workflows.</p> Key Points <p><UnorderedList Mark="Bullet"> <ItemContent> <p>Tumoral components and heterogeneity of oral tongue cancer were investigated on MRI.</p> </ItemContent> <ItemContent> <p>Advanced diffusion-relaxation imaging delineated the tumoral differential profile and predicted metastasis.</p> </ItemContent> <ItemContent> <p>We provided a non-invasive tool for preoperative decision-making in clinical workflows.</p> </ItemContent> </UnorderedList></p> Graphical Abstract <p></p>