Magnetic resonance imaging fractal analysis of O(6)-methylguanine-DNA methyltransferase promoter methylation status in isocitrate dehydrogenase wild-type glioblastoma

Introduction

Glioblastoma (GBM) arises from glial or precursor cells and represents the most aggressive form of primary intracranial malignancy, posing significant challenges in clinical management (1). In 2021, the World Health Organization (WHO) updated the classification of gliomas, underscoring the value of integrating molecular markers with histological classification to enhance diagnosis, treatment strategies, and prognostic predictions (2). Isocitrate dehydrogenase (IDH) wild-type GBM is classified as a WHO grade 4 tumor, associated with a dismal prognosis and a 5-year survival rate of less than 7.1% (3).

O(6)-methylguanine-DNA methyltransferase (MGMT) promoter methylation is a well-established prognostic marker in GBM, as its status influences patient response to temozolomide (TMZ), the standard chemotherapeutic agent for this disease (4). Methylation of the MGMT promoter enhances TMZ sensitivity, improving patient outcomes. Conversely, patients with unmethylated MGMT promoters exhibit resistance to TMZ, often experiencing diminished quality of life due to chemotherapy-associated toxicity without therapeutic benefit (5). While the assessment of MGMT promoter methylation currently requires invasive surgical biopsy, the method’s accuracy is compromised by the spatial and temporal heterogeneity inherent in gliomas. Consequently, there remains an urgent need for non-invasive tools to predict MGMT promoter methylation status in GBM.

Magnetic resonance imaging (MRI) offers a non-invasive approach to brain tumor evaluation and is routinely recommended before initiating treatment. MRI provides hemodynamic, functional, metabolic, microstructural, and genetic insights, facilitating risk stratification and optimizing patient care. Numerous studies have demonstrated associations between MRI parameters and MGMT promoter methylation status in gliomas (6-8). Among the most widely used sequences are T2-weighted imaging (T2WI) and contrast-enhanced T1-weighted imaging (T1C), both of which are critical in clinical practice across healthcare settings (9). However, conventional qualitative analysis cannot fully exploit the wealth of information embedded within these imaging modalities, necessitating the development of more advanced analytical techniques.

Fractal analysis, a mathematical method used to quantify self-similar repeating patterns in non-Euclidean space (10), offers a promising approach for extracting deeper insights from MRI images. Two key parameters in fractal analysis are fractal dimension (FD) and lacunarity. FD measures the morphological complexity of an object (11), while lacunarity reflects the degree of irregularity, inhomogeneity, or translational and rotational invariance within a pattern (12). Fractal analysis has been applied in tumor diagnosis (12,13) and treatment response evaluation (14-16). However, the relationship between MRI-based fractal features and MGMT promoter methylation status in GBM remains unexplored.

In this study, we aim to assess whether fractal analysis of T2WI and T1C sequences can serve as a non-invasive method to predict the methylation status of the MGMT promoter in IDH-wild-type GBM. We present this article in accordance with the TRIPOD reporting checklist (available at https://qims.amegroups.com/article/view/10.21037/qims-2025-726/rc).

Methods

Study design and patients

This retrospective study was approved by the Ethics Committee of Lanzhou University Second Hospital (Project No. 2020A-070), and individual consent for this retrospective analysis was waived. This study was conducted in accordance with the Declaration of Helsinki and its subsequent amendments. A total of 303 patients with GBM confirmed by surgical pathology were included between January 2018 and June 2023. The training set consisted of 220 patients from first center, while the validation cohort comprised 83 patients from second center. All patients met the following inclusion criteria: (I) no prior biopsy or pharmacological treatment of the lesion before MRI; (II) histopathological confirmation of IDH-wild-type GBM; and (III) age ≥18 years. Exclusion criteria were: (I) image quality compromised by motion artifacts, poor contrast enhancement, missing sequences, signal inhomogeneity, or incomplete tumor coverage; (II) missing data on MGMT promoter methylation status; and (III) history of brain surgery or traumatic brain injury. A flowchart illustrating the study design is presented in Figure 1.

Figure 1 Study design and analysis. Workflow of image pre-processing, feature extraction and selection, model building, and interpretation. DCA, decision curve analysis; MGMT, O(6)-methylguanine-DNA methyltransferase; ROC, receiver operating characteristic; T1C, contrast-enhanced T1-weighted imaging; T2WI, T2-weighted imaging.

MRI protocol

All MRI scans were performed using a Siemens Verio 3.0 T (Center 1) and a 1.5 Tesla GE Signa scanner (Center 2) superconducting scanner, with patients positioned supine. The Siemens Verio 3.0 T (Center 1) protocol included: T1-weighted imaging (T1WI): gradient echo (GRE) sequence, repetition time (TR) 550 ms, echedelay time (TE) 11 ms, slice thickness 5.0 mm, inter-slice gap 1.5 mm, FOV 260 mm
× 260 mm, matrix size 256×256. T2WI: TSE sequence, TR 2,200 ms, TE 96 ms, echo chain length 8, number of excitations 2. The 1.5 Tesla GE Signa scanner (Center 2) protocol included: T1WI: GRE sequence, TR 2,250 ms, TE 24 ms, slice thickness 5.0 mm, inter-slice gap 1.5 mm, FOV 240 mm × 240 mm, matrix size 256×256. T2WI: TSE sequence, TR 5,100 ms, TE 129.48 ms, echo chain length 8, number of excitations 2. Contrast-enhanced scans used gadolinium-DTPA (Gd-DTPA) as the contrast agent (Bayer Schering Pharma AG, Berlin, Germany) administered intravenously at 0.1 mmol/kg with a flow rate of 3.0 mL/s.

Histopathological analysis

All GBM specimens were resected and subjected to histopathological evaluation by an experienced neuropathologist with 15 years of expertise. The patients we included were reclassified as IDH wild-type GBM by a pathologist according to the 2021 WHO CNS tumor classification. MGMT promoter methylation status was determined using methylation-specific polymerase chain reaction as described in the literature (17).

Evaluation of MRI features

Two independent radiologists (with 7 and 9 years of diagnostic neuroradiology experience, respectively) performed a double-blind assessment of routine MRI features for all enrolled patients. Neither radiologist was aware of the MGMT promoter methylation status of the GBM patients. The following features were evaluated: (I) axial maximum tumor diameter; (II) presence or absence of cystic changes; (III) single or multiple lesions; and (IV) presence or absence of hemorrhage.

Fractal analysis

Tumor regions were manually delineated from T2WI and T1C images by two senior neuroradiologists using ITK-SNAP software (version 3.8.0; www.itksnap.org). The largest tumor slice was selected as the region of interest (ROI) for further analysis. For cases with multiple lesions, only the largest lesion was analyzed. FD and lacunarity were extracted using the FracLac plugin for ImageJ (version 2.0.0; http://rsb.info.nih.gov/ij/plugins/fraclac/FLHelp/Introduction.htm) (18). Fractal analysis based on the Box-Counting Method, where the image is divided into a grid to calculate the FD based on the relationship between box size and count. All images were normalized to zero mean and unit variance before ROI analysis. Fractal features were categorized into 26 distinct metrics, including D_B1–D_B8, D_M1–D_M8, D_X1–D_X4, and L₁–L₆. A total of 52 features were derived from the T2WI and T1C sequences, with detailed descriptions available in the literature (12).

Statistical analysis

Inter-observer agreement for conventional MRI features and fractal parameters was assessed using the intraclass correlation coefficient (ICC), with values >0.75 indicating good agreement. Categorical variables were analyzed using chi-square or Fisher’s exact tests, while continuous variables were compared using independent t-tests. Variables with significant differences (P<0.05) were included in multivariate logistic regression to construct predictive models for MGMT promoter methylation status. A nomogram was developed based on logistic regression results to visualize individualized predictions. The model’s performance was evaluated using receiver operating characteristic (ROC) curve analysis, with area under the curve (AUC) values reported alongside 95% confidence intervals (CIs). Youden index was used to determine the optimal cutoff point for model performance. Calibration curves and decision curve analysis (DCA) were used to assess the nomogram’s predictive consistency and clinical utility. All statistical analyses were performed using SPSS (version 25.0; Chicago, IL, USA) and R software (version 2022.02.3).

Results

Comparison of clinical features and conventional MRI features

A total of 220 patients with GBM were included in the study. In the MGMT-methylated group, there were 118 patients (59 males and 59 females) with a mean age of 53.32±12.62 years. The unmethylated group comprised 102 patients (65 males and 37 females) with a mean age of 51.09±10.44 years. A significant difference in gender distribution was observed between the two groups (P<0.05). MRI features demonstrated good inter-observer reliability, with ICC ranging from 0.832 to 0.945. However, no statistically significant differences were identified in other demographic or clinical variables, including tumor diameter, age, presence of Cystic changes, Number of lesions, or Hemorrhage (P>0.05), as shown in Table 1.

Comparison of fractal analysis parameters

Fractal analysis parameters demonstrated good inter-observer reliability, with ICC ranging from 0.756 to 0.917. Significant differences between the methylated and unmethylated groups were found in L₄-T2WI, L₆-T2WI, and L₄-T1C (P<0.05), as summarized in Tables 2,3. Multivariate logistic regression analysis revealed that L₄-T2WI [odds ratio (OR), 0.881; 95% CI: 0.833–0.933; P<0.001] and L₆-T2WI (OR, 1.479; 95% CI: 1.230–1.778; P<0.001) were independent predictors of MGMT promoter methylation status in GBM patients (Table 4).

ROC-curve analysis

A nomogram for predicting MGMT promoter methylation status was developed based on multivariate logistic regression (Figure 2). The model’s predictive performance was evaluated in both the training and validation set. The AUC value for the training set was 0.816 (95% CI: 0.762–0.870), while the validation cohort achieved an AUC of 0.750 (95% CI: 0.644–0.856) (Table 5, Figure 3A,3B). The calibration curves closely aligned with the diagonal line, indicating strong concordance between predicted probabilities and actual outcomes (Figure 4A,4B). DCA demonstrated that the nomogram provided a greater positive net benefit across a wide range of risk thresholds, highlighting its clinical utility and value in decision-making processes (Figure 4C,4D).

Figure 2 The nomogram of predicting the MGMT promoter methylation status in IDH wild-type GBM. GBM, glioblastoma; IDH, isocitrate dehydrogenase; MGMT, O(6)-methylguanine-DNA methyltransferase; T2WI, T2-weighted imaging.

Figure 3 The ROC curve of the nomogram in the training set (A) and validation set (B). ROC, receiver operating characteristic.

Figure 4 Calibration curves predicted by nomogram in the training set (A) and validation set (B). DCA curves predicted by nomogram in the training set (C) and validation set (D). DCA, decision curve analysis.

Discussion

In this study, 303 GBM patients from two centers were analyzed, and fractal analysis identified two key predictive variables for MGMT promoter methylation status: L₄-T2WI and L₆-T2WI. A predictive model built on these two fractal features demonstrated strong performance in distinguishing between methylated and unmethylated MGMT status. This model holds significant clinical potential, offering the opportunity for preoperative identification of patients who may benefit from TMZ chemotherapy. By integrating these predictive insights, clinicians can initiate early and targeted treatment, potentially improving outcomes while reducing the risk of adverse effects associated with ineffective chemotherapy.

Although routine clinical signs and conventional MRI features are often used initially to assess MGMT promoter methylation status, their utility remains limited. Previous studies, such as that by Jang et al. (19), have reported that female patients with GBM tend to exhibit higher MGMT methylation levels and better prognoses than male patients, findings that are consistent with these results. However, other research has shown that conventional MRI features alone do not reliably predict MGMT methylation status (20).
In this study, we similarly found no significant differences in conventional MRI features between patients with methylated and unmethylated MGMT promoters. These results underscore the limitations of conventional MRI in non-invasive preoperative prediction of MGMT methylation status, highlighting the need for alternative approaches.

Fractal analysis presents a promising tool for characterizing tumor texture and complexity in MRI images. This method has been widely used to assess the complexity of pathological patterns (21). Fractal analysis is a method for measuring the complexity of structures and geometric patterns that are relatively stable and less susceptible to imaging noise than other textural features. It has been successfully applied to various glioma studies, including IDH mutation prediction (21), glioma grading (22), and prognostication of GBM patients (14). FD provides a quantitative measure of how efficiently a shape fills space, accounting for heterogeneity across tumor structures. Higher FD values are associated with more fragmented, heterogeneous patterns (14). Lacunarity is defined as the degree of gappiness, inhomogeneity or translational and rotational invariance. Higher lacunarity indicates greater heterogeneity and rotational variance (18). In this study, we observed that the unmethylated MGMT group exhibited significantly higher cleavage values (L₄-T2WI and L₆-T2WI) than the methylated group, the AUC value reached 0.750 in the validation set. The AUC of the model exceeded that of both conventional MRI features and predictive models built using radiomics (23,24). This difference likely reflects the rapid and unregulated proliferation of tumor cells with unmethylated MGMT promoters, leading to increased heterogeneity. In contrast, methylated MGMT promoters are associated with reduced DNA repair capacity and slower tumor cell proliferation (25,26). Methylated tumors have blurrier boundaries and more aggressive tumor invasion than unmethylated GBM, which also makes their shape more irregular (27). This may have led to greater GBM heterogeneity in the MGMT unmethylated group, with higher cleavage values than in the methylated group. FD/lacunarity regarding MGMT methylation status in GBM were previously investigated without significant results (21). This may be because the study used enhanced subcomponent ROIs for fractal analysis, while we used whole tumor ROIs for fractal analysis, ultimately leading to differences in the results of the two studies.

There are several limitations in this study. First, only two MRI sequences (T2WI and T1C) were analyzed, without incorporating additional multimodal imaging metrics such as diffusion tensor imaging (DTI), diffusion kurtosis imaging (DKI), or perfusion-weighted imaging (PWI). Second, tumor segmentation remains a critical challenge in imaging research. Although automated segmentation methods are advancing, manual segmentation continues to be regarded as more reliable, introducing variability into the analysis. Third, while the predictive model performed well, its diagnostic performance could potentially be enhanced through integration with artificial intelligence methods or further refinement in future studies. Finally, in this study, hemorrhage was based solely on T1WI and T2WI. Susceptibility-weighted imaging (SWI), which is more sensitive for detecting hemorrhage, was not used.

Conclusions

In conclusion, the nomogram developed using fractal analysis provides a practical and effective non-invasive tool for preoperative prediction of MGMT promoter methylation status in GBM patients. This model has the potential to inform personalized therapeutic decisions, guiding the use of TMZ and optimizing clinical outcomes.

Acknowledgments

We thank all the participants for their cooperation in this study.

Footnote

Reporting Checklist: The authors have completed the TRIPOD reporting checklist. Available at https://qims.amegroups.com/article/view/10.21037/qims-2025-726/rc

Data Sharing Statement: Available at https://qims.amegroups.com/article/view/10.21037/qims-2025-726/dss

Funding: This work was supported by the Cuiying Science and Technology Innovation Program of Lanzhou University Second Hospital (grant No. CY2023-YB-A03).

Conflicts of Interest: All authors have completed the ICMJE uniform disclosure form (available at https://qims.amegroups.com/article/view/10.21037/qims-2025-726/coif). All authors report that this work was supported by the Cuiying Science and Technology Innovation Program of Lanzhou University Second Hospital (grant No. CY2023-YB-A03). The authors have no other conflicts of interest to declare.

Ethical Statement: The authors are accountable for all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved. This retrospective study was approved by the Ethics Committee of Lanzhou University Second Hospital (Project No. 2020A-070), and individual consent for this retrospective analysis was waived. This study was conducted in accordance with the Declaration of Helsinki and its subsequent amendments.

Open Access Statement: This is an Open Access article distributed in accordance with the Creative Commons Attribution-NonCommercial-NoDerivs 4.0 International License (CC BY-NC-ND 4.0), which permits the non-commercial replication and distribution of the article with the strict proviso that no changes or edits are made and the original work is properly cited (including links to both the formal publication through the relevant DOI and the license). See: https://creativecommons.org/licenses/by-nc-nd/4.0/.

References

1. Wang X, Ye J, Gao M, Zhang D, Jiang H, Zhang H, Zhao S, Liu X. Nifuroxazide inhibits the growth of glioblastoma and promotes the infiltration of CD8 T cells to enhance antitumour immunity. Int Immunopharmacol 2023;118:109987. [Crossref] [PubMed]
2. Louis DN, Perry A, Wesseling P, Brat DJ, Cree IA, Figarella-Branger D, Hawkins C, Ng HK, Pfister SM, Reifenberger G, Soffietti R, von Deimling A, Ellison DW. The 2021 WHO Classification of Tumors of the Central Nervous System: a summary. Neuro Oncol 2021;23:1231-51. [Crossref] [PubMed]
3. Ostrom QT, Price M, Neff C, Cioffi G, Waite KA, Kruchko C, Barnholtz-Sloan JS. CBTRUS Statistical Report: Primary Brain and Other Central Nervous System Tumors Diagnosed in the United States in 2015-2019. Neuro Oncol 2022;24:v1-v95. [Crossref] [PubMed]
4. Weller M, Tabatabai G, Kästner B, Felsberg J, Steinbach JP, Wick A, et al. MGMT Promoter Methylation Is a Strong Prognostic Biomarker for Benefit from Dose-Intensified Temozolomide Rechallenge in Progressive Glioblastoma: The DIRECTOR Trial. Clin Cancer Res 2015;21:2057-64. [Crossref] [PubMed]
5. Liu X, Han T, Wang Y, Liu H, Zhou J. Prediction of O(6)-methylguanine-DNA methyltransferase promoter methylation status in IDH-wildtype glioblastoma using MRI histogram analysis. Neurosurg Rev 2024;47:285. [Crossref] [PubMed]
6. Ahn SS, Shin NY, Chang JH, Kim SH, Kim EH, Kim DW, Lee SK. Prediction of methylguanine methyltransferase promoter methylation in glioblastoma using dynamic contrast-enhanced magnetic resonance and diffusion tensor imaging. J Neurosurg 2014;121:367-73. [Crossref] [PubMed]
7. Bahrami N, Hartman SJ, Chang YH, Delfanti R, White NS, Karunamuni R, Seibert TM, Dale AM, Hattangadi-Gluth JA, Piccioni D, Farid N, McDonald CR. Molecular classification of patients with grade II/III glioma using quantitative MRI characteristics. J Neurooncol 2018;139:633-42. [Crossref] [PubMed]
8. Han Y, Yan LF, Wang XB, Sun YZ, Zhang X, Liu ZC, Nan HY, Hu YC, Yang Y, Zhang J, Yu Y, Sun Q, Tian Q, Hu B, Xiao G, Wang W, Cui GB. Structural and advanced imaging in predicting MGMT promoter methylation of primary glioblastoma: a region of interest based analysis. BMC Cancer 2018;18:215. [Crossref] [PubMed]
9. Xue C, Zhou Q, Zhang P, Zhang B, Sun Q, Li S, Deng J, Liu X, Zhou J. MRI histogram analysis of tumor-infiltrating CD8+ T cell levels in patients with glioblastoma. Neuroimage Clin 2023;37:103353. [Crossref] [PubMed]
10. Lookian PP, Chen EX, Elhers LD, Ellis DG, Juneau P, Wagoner J, Aizenberg MR. The Association of Fractal Dimension with Vascularity and Clinical Outcomes in Glioblastoma. World Neurosurg 2022;166:e44-51. [Crossref] [PubMed]
11. Fernández E, Jelinek HF. Use of fractal theory in neuroscience: methods, advantages, and potential problems. Methods 2001;24:309-21. [Crossref] [PubMed]
12. Liu S, Wang X, Liu X, Li S, Liao H, Qiu X. Non-invasive differential diagnosis of teratomas from other intracranial germ cell tumours using MRI-based fractal and radiomic analyses. Eur Radiol 2024;34:1434-43. [Crossref] [PubMed]
13. Donato I, Velpula KK, Tsung AJ, Tuszynski JA, Sergi CM. Demystifying neuroblastoma malignancy through fractal dimension, entropy, and lacunarity. Tumori 2023;109:370-8. [Crossref] [PubMed]
14. Curtin L, Whitmire P, White H, Bond KM, Mrugala MM, Hu LS, Swanson KR. Shape matters: morphological metrics of glioblastoma imaging abnormalities as biomarkers of prognosis. Sci Rep 2021;11:23202. [Crossref] [PubMed]
15. Tochigi T, Kamran SC, Parakh A, Noda Y, Ganeshan B, Blaszkowsky LS, Ryan DP, Allen JN, Berger DL, Wo JY, Hong TS, Kambadakone A. Response prediction of neoadjuvant chemoradiation therapy in locally advanced rectal cancer using CT-based fractal dimension analysis. Eur Radiol 2022;32:2426-36. [Crossref] [PubMed]
16. Kurata Y, Hayano K, Ohira G, Imanishi S, Tochigi T, Isozaki T, Aoyagi T, Matsubara H. Computed tomography-derived biomarker for predicting the treatment response to neoadjuvant chemoradiotherapy of rectal cancer. Int J Clin Oncol 2021;26:2246-54. [Crossref] [PubMed]
17. Paz MF, Yaya-Tur R, Rojas-Marcos I, Reynes G, Pollan M, Aguirre-Cruz L, García-Lopez JL, Piquer J, Safont MJ, Balaña C, Sanchez-Cespedes M, García-Villanueva M, Arribas L, Esteller M. CpG island hypermethylation of the DNA repair enzyme methyltransferase predicts response to temozolomide in primary gliomas. Clin Cancer Res 2004;10:4933-8. [Crossref] [PubMed]
18. Liu S, Fan X, Zhang C, Wang Z, Li S, Wang Y, Qiu X, Jiang T. MR imaging based fractal analysis for differentiating primary CNS lymphoma and glioblastoma. Eur Radiol 2019;29:1348-54. [Crossref] [PubMed]
19. Jang B, Yoon D, Lee JY, Kim J, Hong J, Koo H, Sa JK. Integrative multi-omics characterization reveals sex differences in glioblastoma. Biol Sex Differ 2024;15:23. [Crossref] [PubMed]
20. Gupta A, Omuro AM, Shah AD, Graber JJ, Shi W, Zhang Z, Young RJ. Continuing the search for MR imaging biomarkers for MGMT promoter methylation status: conventional and perfusion MRI revisited. Neuroradiology 2012;54:641-3. [Crossref] [PubMed]
21. Yadav N, Mohanty A, V A, Tiwari V. Fractal dimension and lacunarity measures of glioma subcomponents are discriminative of the grade of gliomas and IDH status. NMR Biomed 2024;37:e5272. [Crossref] [PubMed]
22. Battalapalli D, Vidyadharan S, Prabhakar Rao BVVSN, Yogeeswari P, Kesavadas C, Rajagopalan V. Fractal dimension: analyzing its potential as a neuroimaging biomarker for brain tumor diagnosis using machine learning. Front Physiol 2023;14:1201617. [Crossref] [PubMed]
23. Verduin M, Primakov S, Compter I, Woodruff HC, van Kuijk SMJ, Ramaekers BLT, et al. Prognostic and Predictive Value of Integrated Qualitative and Quantitative Magnetic Resonance Imaging Analysis in Glioblastoma. Cancers (Basel) 2021;13:722. [Crossref] [PubMed]
24. Setyawan NH, Choridah L, Nugroho HA, Malueka RG, Dwianingsih EK, Supriatna Y, Supriyadi B, Hartanto RA. Glioma Grade and Molecular Markers: Comparing Machine-Learning Approaches Using VASARI (Visually AcceSAble Rembrandt Images) Radiological Assessment. Cureus 2024;16:e63873. [Crossref] [PubMed]
25. Chen X, Zhang M, Gan H, Wang H, Lee JH, Fang D, Kitange GJ, He L, Hu Z, Parney IF, Meyer FB, Giannini C, Sarkaria JN, Zhang Z. A novel enhancer regulates MGMT expression and promotes temozolomide resistance in glioblastoma. Nat Commun 2018;9:2949. [Crossref] [PubMed]
26. You SH, Choi SH, Kim TM, Park CK, Park SH, Won JK, Kim IH, Lee ST, Choi HJ, Yoo RE, Kang KM, Yun TJ, Kim JH, Sohn CH. Differentiation of High-Grade from Low-Grade Astrocytoma: Improvement in Diagnostic Accuracy and Reliability of Pharmacokinetic Parameters from DCE MR Imaging by Using Arterial Input Functions Obtained from DSC MR Imaging. Radiology 2018;286:981-91. [Crossref] [PubMed]
27. Noushmehr H, Weisenberger DJ, Diefes K, Phillips HS, Pujara K, Berman BP, et al. Identification of a CpG island methylator phenotype that defines a distinct subgroup of glioma. Cancer Cell 2010;17:510-22. [Crossref] [PubMed]