The adult-type diffuse gliomas are the most common type of malignant brain tumor in adults [1]. The 2016 World Health Organization Classification of Central Nervous System Tumors (WHO CNS 4) [2] introduced molecular profiling as a diagnostic criterion for gliomas for the first time, emphasizing its clinical significance. The 2021 revision (WHO CNS 5) [3] further expanded the role of molecular classification by incorporating additional key molecular markers. For example, the deletion of cyclin-dependent kinase inhibitor 2 A/B (CDKN2A/B) is now recognized as a defining feature that upgrades IDH-mutant lower-grade astrocytoma to WHO grade 4 astrocytoma. Similarly, mutations in the telomerase reverse transcriptase (TERT) promoter, epidermal growth factor receptor (EGFR) gene amplification, concurrent gain of the whole chromosome 7 and loss of the whole chromosome 10 (+7/−10) are considered molecular hallmarks for upgrading IDH-wildtype lower-grade gliomas to glioblastoma.

These updates, grounded in molecular genetics, suggest that molecularly defined grade 4 gliomas share similar therapeutic characteristics with histologically defined grade 4 gliomas. Historically, histological grade 4 gliomas exhibit microvascular proliferation or necrosis, which typically appear as ring enhancements or necrotic areas on imaging. In contrast, molecularly upgraded grade 4 gliomas often lack these typical radiographic features, making accurate grading based on imaging alone more difficult [4].

To date, only studies utilizing ADC maps have attempted to predict WHO CNS 5 grading, yielding better results than those predicting histological grading [5]. Quantitative susceptibility mapping (QSM), an advancement of SWI, can calculate the content of substances such as iron, hemosiderin, de-oxyhemoglobin and calcification in tissue [[6], [7], [8]]. Therefore, QSM can identify calcifications, microbleeds, and microvessels in gliomas. In recent years, several studies have leveraged QSM to predict the histological grading and molecular subtypes of gliomas [[9], [10], [11]]. For instance, Wenting Rui et al. demonstrated strong performance in predicting the histological grading and molecular subtypes of gliomas using a deep learning-assisted QSM approach [9]. Diffusion kurtosis imaging (DKI) is based on the non-Gaussian diffusion model of water molecules, which offers a more accurate depiction of tissue characteristics than diffusion tensor imaging (DTI) and diffusion-weighted imaging (DWI) [[12], [13], [14], [15]]. Previous studies have shown the efficacy of DKI in the histological grading and molecular classification of gliomas. For example, Ankang Gao et al. achieved good results in predicting glioma grading and molecular classification using DKI histograms [16]. Similarly, other studies have reached similar conclusions [[17], [18], [19], [20], [21]]. However, both QSM and DKI have mainly been utilized in studies focused on histological grading. Research based on molecular grading according to the WHO CNS 5 classification remains limited.

We hypothesize that the combination of histogram features from both DKI and QSM will add value to conventional sequences in predicting molecular grading. Therefore, this study uses histogram features of QSM, DKI, and conventional sequences to predict WHO CNS 5 grading for adult-type diffuse gliomas, and explores the survival differences between gliomas of different grades to provide more insights for imaging biomarkers in clinical treatment decision-making.