Glioma is the most common primary tumor of the central nervous system (CNS), accounting for 80% ∼ 85% of all malignant brain tumors in adults [1], [2]. The 2021 World Health Organization (WHO) classification of CNS tumors integrates molecular profiling with histopathology for the precise classification of diffuse glioma [3]. Among these molecular phenotypes, isocitrate dehydrogenase (IDH) mutation status serves as a pivotal biomarker, with critical implications for accurate diagnosis and therapeutic decision-making [4]. Compared with IDH wild-type (IDHwt) gliomas, IDH mutant (IDHmut) gliomas exhibit a more favorable prognosis [5], [6].

Currently, determining IDH mutation status in gliomas relies on invasive biopsy or surgical resection, which carries sampling errors due to spatial heterogeneity [7]. In contrast, magnetic resonance imaging (MRI) enables non-invasive whole-tumor evaluation and is fundamental to glioma management, facilitating diagnosis, treatment planning, and prognostication [8], [9], [10]. Notably, gliomas exhibiting ring enhancement with central necrosis on MRI are often indicative of IDHwt subtype. In contrast, non-enhancing gliomas comprise both IDHwt and IDHmut subtypes with profoundly different prognoses [11], [12]. Therefore, non-invasive determination of IDH mutation status in non-enhancing gliomas prior to surgery is crucial for informing clinical management and prognostic stratification.

In clinical practice, evaluations of IDH mutation status in gliomas primarily rely on semantic features derived from conventional MRI [12], [13]. IDHmut gliomas are predominantly located in the frontal lobe and typically demonstrate well-defined margins, homogeneous signal intensity, and infrequent necrosis [14], [15]. In contrast, IDHwt gliomas are characterized by multilobar involvement, ill-defined margins, and frequent necrosis [16]. In addition, T2-FLAIR mismatch sign is a 100% specificity for IDH-mutant astrocytomas [17], [18]. Although highly interpretable and practical for clinical application, the assessment of these semantic features depends critically on radiologists' subjective judgment, rendering them vulnerable to inter- and intra-observer variability. Hence, integrating semantic features with objective quantitative parameters holds significant promise for improving the accuracy of IDH mutation status prediction.

Time-dependent diffusion MRI (TDD-MRI) has emerged as a valuable approach for the non-invasive characterization of tumor microstructure [19], [20], thus compensating for the inherent quantitative limitations of semantic feature. Conventional diffusion kurtosis imaging (DKI) or diffusion tensor imaging (DTI) probe anisotropic water diffusion at the millimeter scale by employing multiple gradient directions [21], [22], [23]. In contrast, oscillating gradient spin-echo (OGSE) sequences utilize high-frequency gradients to characterize tissue at the micrometer scale, offering enhanced sensitivity to cellular dimensions and greater specificity in resolving the tissue microenvironment. By combining OGSE and pulsed gradient spin-echo (PGSE) sequences, TDD-MRI quantifies key microstructural parameters relevant to tumor pathology, including cell diameter, cell density, intracellular volume fraction, and extracellular diffusion coefficient [24], [25]. TDD-MRI has demonstrated promise for risk stratification in cancers of the breast, prostate, endometrium and pediatric glioma [25], [26], [27], [28]. However, the potential of TDD-MRI for IDH mutation prediction, particularly in non-enhancing gliomas, remains unexplored.

This study aims to combine semantic features with microstructural parameters derived from TDD-MRI to predict IDH mutation status in non-enhancing gliomas.