Continuous-time random walk and fractional order calculus models histogram analysis of glioma biomarkers, including IDH1, ATRX, MGMT, and TERT, on differentiation

Introduction

The fifth edition of World Health Organization (WHO) classification of central nervous system (CNS) tumors (WHO CNS5) emphasized the importance of classifying tumors according to their molecular subtypes. The WHO CNS5 has formally incorporated the isocitrate dehydrogenase (IDH) mutation status into the disease nomenclature due to the significant genetic and epigenetic differences between IDH-mutant (IDH-Mut) and IDH-wildtype (IDH-WT) gliomas (1). The X-linked alpha-thalassemia/mental retardation syndrome (ATRX) mutation (loss of ATRX protein expression, ATRX−) indicates a better prognosis for patients with IDH-mutant astrocytomas, whereas ATRX-WT (ATRX+) has been found to correlate with tumor recurrence (2-5). The O6-methylguanine-DNA-methyltransferase (MGMT) is a DNA repair enzyme that has been reported as a predictive biomarker for the response to alkylating chemotherapy (6), and glioma patients with MGMT promoter methylation (MGMTp-met) have favorable clinical outcomes (7,8). Mutations in the telomerase reverse transcriptase promoter (TERTp-Mut) lead to increased telomerase activity and telomere lengthening, which are associated with glioma aggressiveness (9). The IDH-WT diffuse gliomas with TERTp-Mut have been reported to have the least favorable overall survival (OS) (10,11). Early identification of the molecular biomarkers such as IDH1, ATRX, MGMT, and TERT can provide valuable diagnostic and prognostic information. However, genetic typing of gliomas through immunohistochemistry or sequencing is an invasive, expensive method and susceptible to sample bias (12). Given these challenges, preoperative non-invasive magnetic resonance imaging (MRI) becomes important for preliminary molecular genotyping of gliomas and clinical diagnosis.

The apparent diffusion coefficient (ADC) value is derived from diffusion-weighted imaging (DWI) based on a Gaussian diffusion model, which assumes uniform intravoxel diffusion (13). However, the microenvironment in biological tissues is more complex, especially with fibers, vascular walls, and cell membrane restrictions in heterogeneous tumors, leading to non-Gaussian diffusion (14-16). To overcome the limitations of DWI, various non-Gaussian diffusion models have been developed to extract information on tissue microstructure, better reflecting tumor heterogeneity (17-20). Among these, the high b-value diffusion MRI technique based on the continuous-time random walk (CTRW) model (21,22) and its predecessor, the fractional order calculus (FROC) model (23), have been developed. The CTRW model considers that the movement of water molecules may take variable times (water trapping) or travel variable distances (water jumping), reflecting temporal and spatial diffusion heterogeneity through parameters α_CTRW and β_CTRW, respectively. The above two parameters, along with anomalous diffusion coefficient D_CTRW, enable the CTRW model to provide a comprehensive assessment of tissue changes (21,24,25). The FROC model generates a number of parameters, including the diffusion coefficient D_FROC, fractional-order derivative in space β_FROC, and the spatial parameter µ_FROC which is proposed as a measure of the diffusion mean free length. These parameters improve tumor characterization by describing the diffusion process (D_FROC) and intravoxel tissue heterogeneity (β_FROC) (25-27).

Recently, many studies have used MRI combined with histogram analysis to assess tumor grading and molecular type, demonstrating promising performance (28,29). Histogram analysis has been shown to have several advantages. First, it can provide more statistical information than the mean value alone and offer a quantitative mechanism for analyzing non-significant changes in tumor voxels. Second, the percentiles may help to assess the malignant microenvironments. In addition, changes in asymmetry and shape of the histogram, represented by skewness and kurtosis, may indicate microstructural changes (30,31).

We hypothesized that the CTRW and FROC models may reveal differences in water diffusivity and intravoxel heterogeneity in tumor tissues between different molecular subtypes of diffuse gliomas. Therefore, the goal of this study was to demonstrate the feasibility of using the CTRW and FROC models, together with the histogram analysis, to identify the molecular subtypes (IDH1, ATRX, MGMT, and TERT) of diffuse gliomas in adults. We present this article in accordance with the STARD reporting checklist (available at https://qims.amegroups.com/article/view/10.21037/qims-2024-2725/rc).

Methods

Patient population

This study was approved by the Institutional Review Board of Tongji Hospital of Tongji Medical College of Huazhong University of Science and Technology (No. TJ-IRB202412200) and written informed consent was provided by all participants. This study was conducted in accordance with the Declaration of Helsinki and its subsequent amendments. The image data of consecutive patients admitted to our hospital from March 2017 to January 2022 were retrospectively analyzed. The inclusion criteria were as follows: (I) pathologically confirmed primary gliomas; (II) known IDH1 genotype, and/or ATRX mutation status, MGMT promoter status, and TERT promoter status; and (III) underwent MRI before surgery or biopsy. The exclusion criteria were as follows: (I) age less than 18 years; (II) purely cystic gliomas; (III) MR images with artifacts or poor quality; and (IV) lack of routine MR images. One case younger than 18 years, two cases with purely cystic glioma, three cases with severe motion artifacts, and two cases lacking routine MRI images were excluded. A total of 111 patients were recruited for analysis. The flowchart for participant selection criteria is shown in Figure 1.

Figure 1 Flowchart showing participant selection criteria. MRI, magnetic resonance imaging; IDH1, isocitrate dehydrogenase 1.

Image data acquisition

All images were performed on the a 3T MR system (Discovery MR750, GE Medical Systems, Milwaukee, WI, USA) with a 32-channel head coil. Routine axial sequences included T1 fluid-attenuated inversion recovery (T1-FLAIR), T2 fast spin-echo (T2-FSE), T2-FLAIR, and contrast-enhanced T1-weighted (CE-T1W).

Multi-b-value DWI was obtained using spin-echo echo-planar imaging sequences before the injection of contrast agents and performed with 20 b-values (b=0, 20, 50, 80, 100, 150, 200, 400, 600, 800, 1,000, 1,200, 1,500, 2,000, 2,400, 2,800, 3,200, 3,600, 4,000, and 4,500 s/mm², with non-zero b-values applied in three orthogonal diffusion directions), separation between two diffusion gradient lobes Δ=42.8 ms, duration of each diffusion gradient lobe δ=28.2 ms, for 0–1,000 s/mm², number of excitations (NEX) =1, for 1,200–2,800 s/mm², NEX =2, and for 3,200–4,500 s/mm², NEX = 4, repetition time/echo time (TR/TE) =3,200/90.6 ms, matrix =160×160, slice thickness =5 mm, spacing =1.5 mm, field of view (FOV) =240×240 mm², acquisition time was 5 minutes 52 seconds.

Acquisition of molecular genotypes

The IDH1 genotype was evaluated using immunohistochemical staining or next-generation sequencing. ATRX mutation status in 56 patients was detected by immunohistochemistry. MGMT promoter methylation status in 78 patients was performed using pyrophosphate sequencing, with a methylation level exceeding 10% at four CpG sites on chromosome 10 considered positive. TERT promoter mutation status in 74 patients was determined using multiplex polymerase chain reaction (PCR) combined with next-generation sequencing, and C228T and C250T mutations were analyzed.

Image processing

The mono-exponential, CTRW, and FROC models were calculated using a software prototype developed for MIWP-MRstation (Chengdu Zhongying Medical Technology Co., Chengdu, Sichuan, China).

The ADC was obtained by the following mono-exponential model:

S/S0=exp(−b⋅ADC)[1]

where S₀ and S represent the signal intensities at b-values of 0 and 1,000 s/mm², respectively.

Three diffusion parameters (D_CTRW, α_CTRW, and β_CTRW) was calculated according to a simplified mathematical equation:

S/S0=Eα[−(bD)β][2]

where S₀ is the signal intensity without diffusion weighting. Eα yields a characteristic decay process represented by a Mittag-Leffler function. D (in ×10^-3mm²/s) is the anomalous diffusion coefficient. α (unitless; 0<α≤1) is the temporal diffusion heterogeneity index, and β (unitless; 1/2<β≤1) is the spatial diffusion heterogeneity index. In addition to analyzing the CTRW model, we further computed the parameter maps for three special cases: sub-diffusion (0<α≤1, β=1), super-diffusion (α=1, 1/2 <β≤1), and quasi-diffusion (α=β). The CTRW model and three special cases are unified under the same CTRW framework (32,33). The detailed methodology for sub-, super-, and quasi-diffusion models is provided in the Supplementary file (Appendix 1).

Three diffusion parameters (D_FROC, β_FROC, and µ_FROC) were calculated by a complicated formula:

S=S0exp[−Dμ2(β−1)(γGdδ)2β(Δ−2β−12β+1δ)][3]

where S₀ has the same meanings as above. G_d is the diffusion gradient amplitude, δ is the duration of each diffusion gradient lobe, and Δ is the gradient lobe separation. D (in ×10⁻³ mm²/s) is the diffusion coefficient. β (unitless; 0<β≤1) is the fractional-order derivative in space; and µ (in µm) is a spatial parameter. Multi-b-value DWI were fitted to the FROC diffusion model voxel by voxel by using a Levenberg-Marquardt nonlinear fitting algorithm. In the fitting, D was estimated by a mono-exponential model with data acquired at lower b values (<2,000 s/mm²). After D was calculated, β and µ were then obtained from the voxel-wise nonlinear fitting algorithm using all b values.

Regions of interest (ROIs) analysis

Two neuroradiologists, blinded to the glioma subtypes, independently delineated the ROIs. All parameter maps and routine MR images were analyzed using ITK-SNAP software (https://www.itksnap.org/pmwiki/pmwiki.php). Referring to routine sequences, ROIs contained the whole solid parts of tumor drawn on each diffusion image with b=0 s/mm², avoiding hemorrhage, calcification, edema, necrosis, and cyst. ROIs were then propagated to other parameter maps, and the histogram features including the 10^th percentile, 90^th percentile, kurtosis, mean, median, and skewness were calculated based on diffusion parameter maps derived from CTRW, FROC and mono-exponential models.

Statistical analysis

The statistical analysis was performed with the software SPSS 26.0 (IBM Corp., Armonk, NY, USA) and MedCalc 18.2 (MedCalc, Ostend, Belgium). The inter-observer reliability was evaluated by calculating the intra-class correlation coefficient (ICC). The distribution normality of continuous variables was assessed by Shapiro-Wilk test. The unpaired t-test (for normally distributed data) and Mann-Whitney U test (for non-normally distributed data) were used to compare histogram parameters derived from the CTRW (D, α, β), FROC (D, β, µ), and mono-exponential (ADC) models between two groups regarding molecular subtypes (IDH1-mutant vs. IDH1-WT, IDH1+/ATRX+ vs. IDH1+/ATRX−, MGMT promoter methylation vs. MGMT promoter unmethylation, and IDH1-WT/TERTp-Mut vs. Non-IDH1-WT/TERTp-Mut). Chi-squared tests were used to test categorical variables. Receiver operating characteristic (ROC) curves were employed to calculate the area under the curve (AUC), sensitivity, specificity, and cut-off value. The AUCs were compared using DeLong test. The false discovery rate (FDR, Benjamini-Hochberg method) was used for multiple comparisons. To investigate whether combining parameters within each diffusion model could provide better diagnostic performance, a binary forward stepwise logistic regression analysis was conducted. A P value <0.05 was taken to indicate statistical significance.

Results

Clinical characteristics

This retrospective study enrolled 111 patients with diffuse gliomas (62 males and 49 females, age 48.14±11.59 years). Based on the molecular phenotype findings of IDH1, ATRX, MGMT, and TERT, patients were divided into four comparative experimental groups. Except for the MGMT promoter status in gliomas, the above groups were significantly influenced by the age of the patients. In addition, the gender showed no significant difference between different groups of each molecular subtype (Table 1). Figure 2 shows the CE-T1W image and diffusion parameter maps of representative cases of IDH1-mutant and IDH1-WT glioma patients. Figure 3 shows the images of typical IDH1+/ATRX+ and IDH1+/ATRX− glioma patients.

Figure 2 Representative images from patients with IDH1-mutant and IDH1-wildtype diffuse gliomas. (A₁-H₁) Representative images from a 38-year-old female with IDH1-mutant glioma (WHO grade 4). The images include CE-T1W image (A₁), ADC image (B₁), D_CTRW image (C₁), α_CTRW image (D₁), and β_CTRW image (E₁); D_FROC image (F₁), β_FROC image (G₁), and µ_FROC image (H₁). (A₂-H₂) Representative images from a 54-year-old male patient with IDH1-wildtype glioma (WHO grade 4). The images include CE-T1W image (A₂), ADC image (B₂), D_CTRW image (C₂), α_CTRW image (D₂), and β_CTRW image (E₂); D_FROC image (F₂), β_FROC image (G₂), and µ_FROC image (H₂). ADC, apparent diffusion coefficient; CE-T1W, contrast-enhanced T1-weighted; CTRW, continuous-time random walk; FROC, fractional order calculus; IDH1, isocitrate dehydrogenase 1; WHO, World Health Organization.

Figure 3 Representative images from IDH1-mutant glioma patients with different ATRX status. (A₁-H₁) Representative images from a 60-year-old female with IDH1+/ATRX+ diffuse glioma (WHO grade 2). The images include CE-T1W image (A₁), ADC image (B₁), D_CTRW image (C₁), α_CTRW image (D₁), and β_CTRW image (E₁); D_FROC image (F1), β_FROC image (G₁), and µ_FROC image (H₁). (A₂-H₂) Representative images from a 39-year-old male patient with IDH1+/ATRX− diffuse glioma (WHO grade 2). The images include CE-T1W image (A₂), ADC image (B₂), D_CTRW image (C₂), α_CTRW image (D₂), and β_CTRW image (E₂); D_FROC image (F₂), β_FROC image (G₂), and µ_FROC image (H₂). ADC, apparent diffusion coefficient; ATRX, X-linked alpha-thalassemia/mental retardation syndrome; ATRX+, ATRX wildtype; ATRX−, ATRX-mutant; CE-T1W, contrast-enhanced T1-weighted; CTRW, continuous-time random walk; FROC, fractional order calculus; IDH1, isocitrate dehydrogenase 1; IDH1+, IDH1-mutant; WHO, World Health Organization.

The 10^th percentile, 90^th percentile, mean, and median of mono-exponential, CTRW, and FROC parameters, exhibited good inter-observer reliability (ICCs: 0.921–0.982). However, the ICCs of the kurtosis and skewness derived from three diffusion models ranged from 0.347 to 0.837. For parameters with an ICC below 0.75, the results obtained from the more experienced neuroradiologist were used in the analysis instead.

Statistical analysis in IDH1 genotyping

In IDH1 genotyping, statistically significant differences (P<0.05) were found between IDH1-mutant and wildtype gliomas in all histogram parameters except for the kurtosis D_CTRW, 90^th percentile α_CTRW, kurtosis β_FROC, 90^th percentile µ_FROC, and kurtosis ADC after FDR correction (Table 2). Figure 4A-4D presents box and whisker plots of the four histogram parameters (10^th percentile β_CTRW, 10^th percentile β_FROC, mean β_FROC and mean β_CTRW) with the best diagnostic performance, illustrating that these parameters were significantly higher in the IDH1-mutant group than those in IDH1-WT group. The ROC analysis of the IDH1 genotype (Table 3, Figure 4E) showed that the 10^th percentile β_CTRW (AUC =0.854), the 10^th percentile β_FROC (AUC =0.853), the mean β_FROC (AUC =0.843), and the mean β_CTRW (AUC =0.837) were higher than other histogram features. Moreover, the DeLong test (Table S1) revealed that the AUC values of the 10^th percentile β_CTRW, as well as the 10^th percentile, mean, and median β_FROC, were significantly higher than those of all histogram parameters of ADC. In the stepwise logistic regression analysis for each diffusion model (Table S2), the combination of the 10^th percentile β and the median α from the CTRW model improved diagnostic performance, achieving an AUC of 0.866, sensitivity of 92.98%, and specificity of 70.37%.

Figure 4 Group comparisons and ROC curves of the top four histogram parameters derived from the mono-exponential, CTRW, and FROC models in IDH1 genotyping. (A) Comparison of the 10^th percentile β_CTRW, (B) comparison of the 10^th percentile β_FROC, (C) comparison of the mean β_FROC, and (D) comparison of the mean β_CTRW. (E) ROC curves and AUCs of the top four histogram parameters. ****, P<0.0001. AUC, area under the curve; CTRW, continuous-time random walk; FROC, fractional order calculus; IDH1, isocitrate dehydrogenase 1; IDH1-Mut, IDH1-mutant; IDH1-WT, IDH1 wildtype; ROC, receiver operating characteristic.

Statistical analysis for identification of ATRX mutation status in IDH1-mutant gliomas

In the analysis of ATRX genotyping, the diffusion parameters, involving the skewness D_FROC and µ_FROC, as well as 10^th percentile, 90^th percentile, mean, median D_CTRW, D_FROC, µ_FROC, and ADC, had significant differences (P<0.05) between the IDH1+/ATRX− and IDH1+/ATRX+ groups (Table 4). Figure 5A-5D shows that the top four parameters (the 90^th percentile ADC, 90^th percentile D_CTRW, median µ_FROC, and 90^th percentile D_FROC) in the IDH1+/ATRX− group were significantly higher than those in the IDH1+/ATRX+ group. Among IDH1-mutant glioma patients, the ROC analysis of ATRX mutation status (Table 5, Figure 5E) showed that the AUCs of the 90^th percentile ADC, 90^th percentile D_CTRW, median µ_FROC, and 90^th percentile D_FROC were 0.797, 0.790, 0.787, and 0.787, respectively, which were higher than those of other histogram features. The stepwise logistic regression analysis revealed that the FROC model retained 10^th percentile D, 90^th percentile D, 10^th percentile µ, and mean µ, achieving an AUC of 0.853, sensitivity of 76.67%, and specificity of 88.46% (Table S2).

Figure 5 Group comparisons and ROC curves of the top four histogram parameters derived from the mono-exponential, CTRW, and FROC models in distinguishing ATRX mutation status in IDH1-mutant gliomas. (A) Comparison of the 90^th percentile ADC, (B) comparison of the 90^th percentile D_CTRW, (C) comparison of the median µ_FROC, and (D) comparison of the 90^th percentile D_FROC. (E) ROC curves and AUCs of the top four histogram parameters. ***, P<0.001; ****, P<0.0001. AUC, area under the curve; ATRX, X-linked alpha-thalassemia/mental retardation syndrome; ATRX−, ATRX-mutant; ATRX+, ATRX wildtype; CTRW, continuous-time random walk; FROC, fractional order calculus; IDH1, isocitrate dehydrogenase 1; IDH1+, IDH1-mutant; ROC, receiver operating characteristic.

Statistical analysis for identification of MGMT promoter methylation status

Table S3 shows that only the histogram parameters from β_CTRW and β_FROC exhibited significant differences (P<0.05) between the MGMT promoter methylation and unmethylation groups. The ROC analysis showed that the AUCs of the top four parameters (the median, mean, 10^th percentile β_CTRW, and mean β_FROC) were 0.758, 0.755, 0.746, and 0.705, respectively. The diagnostic efficiency of the median β_CTRW was optimal, with 75.61% sensitivity and 67.57% specificity, and the value of median β_CTRW was significantly higher in the MGMT promoter methylation group (Table S4, Figure S1).

Statistical analysis for discriminating IDH1-WT/TERTp-Mut and non-IDH1-WT/TERTp-Mut in diffuse gliomas

Table S5 shows that all histogram parameters from β_CTRW and 10^th percentile, mean, and median β_FROC were significantly different (P<0.05) between IDH1-WT/TERTp-Mut and non-IDH1-WT/TERTp-Mut groups. Figure S2A-S2D demonstrates that the top four parameters, including the 10^th percentile β_CTRW, mean β_CTRW, 10^th percentile β_FROC, and median β_CTRW, exhibited significantly lower values in IDH1-WT/TERTp-Mut group compared to those in the other group. ROC analysis (Table S6, Figure S2E) showed that the AUCs of the 10^th percentile β_CTRW, mean β_CTRW, 10^th percentile β_FROC, and median β_CTRW were 0.869, 0.856, 0.854, and 0.833, respectively.

The detailed results for sub-, super-, and quasi-diffusion models are provided in the Supplementary file (Appendix 1, Tables S7-S14, Figure S3).

Discussion

Our study demonstrated that histogram analysis based on CTRW and FROC models possesses good efficacy for distinguishing molecular subtypes of IDH1, ATRX, MGMT, and TERT in diffuse gliomas. In the analysis of IDH1 genotype and TERT promoter status, the 10^th percentile β_CTRW achieved the best diagnostic performance. For the identification of ATRX mutation status in IDH1-mutant gliomas, the 90^th percentile ADC had the best diagnostic performance, and for the identification of MGMT promoter methylation status, the median β_CTRW achieved the highest AUC value. Thus, we propose that the β_CTRW is a potential non-invasive imaging biomarker for identifying the status of IDH1, MGMT promoter, and TERT promoter in diffuse gliomas.

The ADC derived from the mono-exponential model is limited in characterizing complex tissue microstructures and is susceptible to T2 relaxation effects, which may lead to misinterpretation of diffusion restriction (34). In contrast, the CTRW and FROC diffusion models can reveal the distributive diffusion effects of water molecules and recognize the intravoxel heterogeneity both spatially and temporally within tumors. Karaman et al. (21) indicated that the CTRW model exhibited high specificity (83%), diagnostic accuracy (85%), and predictive ability (AUC =0.957) when grading pediatric brain tumors. Sui et al. validated the feasibility of the FROC model in grading pediatric brain tumors (26) and adult-type diffuse gliomas (27). However, there is a scarcity of research exploring the potential of CTRW and FROC models in distinguishing molecular subtypes of diffuse gliomas in adults.

Previous studies have suggested that IDH-WT gliomas possess more complex tissue architecture and richer microvasculature compared to IDH-mutant gliomas (35-37). Cindil et al. (38) and Zhang et al. (39) have demonstrated the efficacy of ADC to distinguish IDH-mutant gliomas from IDH-WT gliomas. Our findings indicated significantly lower values of the mean ADC, D_CTRW, D_FROC, and µ_FROC in IDH1-wild-type gliomas than in IDH1-mutant gliomas, suggesting more restricted diffusion of water molecules, whereas the parameters representing intravoxel heterogeneity, including α_CTRW, β_CTRW, and β_FROC, showed higher temporal-spatial heterogeneity inside IDH1-WT gliomas. Xie et al. (40) discriminated IDH1-mutant from IDH1-WT gliomas using an alternative non-Gaussian diffusion model based on a stretched-exponential formulism, which is similar to the FROC model. However, the stretched-exponential is developed empirically instead of using the fractionalized Fick diffusion equation (23). Therefore, our study may more accurately reflect the water molecule diffusion conditions of different IDH1 genotypes. Notably, our results demonstrated that the β_FROC and β_CTRW exhibited significantly better diagnostic performance than ADC. This may be attributed to the theory that β_FROC and β_CTRW are characteristic of non-Gaussian distribution and directly reflect the underlying tumor microstructural complexity and heterogeneity, whereas the mono-exponential model was constructed according to the Gaussian distribution. The integration of the 10^th percentile β_CTRW and the median α_CTRW provides a more comprehensive characterization of diffusion process, enhancing the diagnostic efficacy in IDH1 genotyping. Specifically, β reflects spatial heterogeneity by characterizing non-Gaussian distribution of diffusion displacement, whereas α represents temporal heterogeneity by describing the likelihood of water molecules being “trapped” or “released” as they diffuse through the tissue microenvironment (21).

ATRX mutation has become a critical marker for differentiating diffuse astrocytomas from oligodendrogliomas without the necessity for 1p/19q co-deletion testing in IDH-mutant gliomas (2,41). Therefore, our investigation specifically focused on patients with IDH1-mutant gliomas for ATRX mutation status analysis. Our findings underscored that the parameters ADC, D_CTRW, D_FROC, and µ_FROC were effective in differentiating ATRX- and ATRX+ of IDH1-Mut gliomas. Yang et al. (12) and Cheng et al. (42) demonstrated that ATRX+ tumors showed lower ADC values than those in ATRX− tumors, indicating higher cell density in ATRX+ tumors, which is in line with our study. Additionally, our analysis indicated that the mean α_CTRW, β_CTRW, and β_FROC in the ATRX- group were marginally higher than those in the ATRX+ group. However, these differences were not statistically significant, potentially attributable to the limited sample size or the exclusive focus on patients with IDH1-mutant gliomas for ATRX mutation status analysis.

Our study found no statistical difference in ADC, D_CTRW, and D_FROC between the MGMT promoter methylation and unmethylation groups, which is consistent with the findings of Yang et al. (12) and Choi et al. (43), who also reported no significant correlation between ADC values and MGMT promoter methylation status in IDH-Mut astrocytomas and glioblastomas, respectively. However, this is in contrast to the result of Romano et al. (44) who observed a higher ADC value in the MGMT promoter methylation group. The different results may be attributed to the difference in grouping approaches and analysis methods. However, β_CTRW and β_FROC, which reflect tissue heterogeneity, demonstrated statistical differences between the two groups. MGMT promoter methylation prevents the repair of damaged and mutated DNA by MGMT proteins, thereby facilitating apoptosis (45). We hypothesized that the downregulation of MGMT protein can prevent intracellular gene mutation and inhibit cell proliferation, thus reducing the heterogeneity of the tumor microenvironment and leading to an increase in β values.

Suppression of TERT expression has been reported to increase cell sensitivity to DNA damage induced by radiation and chemotherapy, positioning it as a target for novel therapeutic strategies (46,47). Our results showed that the IDH1-WT/TERTp-Mut glioma group had lower β_CTRW and β_FROC values. This might be due to the fact that TERT promoter mutations induce angiogenesis and necrosis (48,49), which further decreases tumor homogeneity, as reflected by lower β values. However, diffusion parameters indicating restricted water molecule diffusion, including ADC, D_CTRW, D_FROC, and µ_FROC, showed no statistical differences between the two groups. This is consistent with findings by Yamashita et al. (50) and Park et al. (51) who observed that ADC values could not distinguish TERT promoter status in gliomas.

In predicting MGMT and TERT promoter status, our findings underscore the limitations of the mono-exponential model, likely because ADC primarily reflects cell density without fully capturing the complexity and heterogeneity of tumor microstructure, which are also influenced by microperfusion, edema, necrosis, and hemorrhage (21). The results highlight the importance of using non-Gaussian diffusion models which can more effectively capture the underlying heterogeneity within tumors and provide better diagnostic efficacy for distinguishing MGMT promoter methylation and TERT promoter status in diffuse gliomas.

Our study has several limitations. First, it involved a small sample size and was a single-center study. Future multicenter studies with larger samples are needed for validation. Second, we did not differentiate between IDH-mutant diffuse gliomas and IDH-WT glioblastomas when analyzing MGMT and TERT promoter status. Previous studies have suggested that there may be differences in prognosis of various MGMT and TERT promoter statuses between different groups. Future investigations will enroll more patients for a deeper study in this aspect. Third, we only used CTRW and FROC diffusion models, which are supplementary to conventional MR sequences. Combining them with conventional sequences may improve diagnostic efficacy. Finally, our study lacks a systematic comparison between the CTRW model and its special cases (sub, super-, and quasi-diffusion models). Future studies will explore this comparison across different diseases.

Conclusions

This study reveals that the CTRW and FROC models, together with histogram analysis, demonstrate superior diagnostic performance compared to the mono-exponential model in predicting different molecular subtypes (IDH1, ATRX, MGMT, and TERT) of diffuse gliomas. Moreover, the CTRW and FROC models offer several new parameters for probing tumor structural heterogeneity at a subvoxel level. The comprehensive characterization of gliomas contributes to personalized clinical management.

Acknowledgments

None.

Footnote

Reporting Checklist: The authors have completed the STARD reporting checklist. Available at https://qims.amegroups.com/article/view/10.21037/qims-2024-2725/rc

Data Sharing Statement: Available at https://qims.amegroups.com/article/view/10.21037/qims-2024-2725/dss

Funding: This study has received funding from the National Key Research and Development Program of China (grant No. 2022YFC2406903) and the National Natural Science Foundation of China (grant No. U22A20354).

Conflicts of Interest: All authors have completed the ICMJE uniform disclosure form (available at https://qims.amegroups.com/article/view/10.21037/qims-2024-2725/coif). All authors report that this study has received funding from the National Key Research and Development Program of China (grant No. 2022YFC2406903) and the National Natural Science Foundation of China (grant No. U22A20354). 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 study was approved by the Institutional Review Board of Tongji Hospital of Tongji Medical College of Huazhong University of Science and Technology (No. TJ-IRB202412200) and written informed consent was provided by all participants. 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. 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]
2. Wiestler B, Capper D, Holland-Letz T, Korshunov A, von Deimling A, Pfister SM, Platten M, Weller M, Wick W. ATRX loss refines the classification of anaplastic gliomas and identifies a subgroup of IDH mutant astrocytic tumors with better prognosis. Acta Neuropathol 2013;126:443-51. [Crossref] [PubMed]
3. Johnson BE, Mazor T, Hong C, Barnes M, Aihara K, McLean CY, et al. Mutational analysis reveals the origin and therapy-driven evolution of recurrent glioma. Science 2014;343:189-93. [Crossref] [PubMed]
4. Ogishima T, Tamura K, Kobayashi D, Inaji M, Hayashi S, Tamura R, Nariai T, Ishii K, Maehara T. ATRX status correlates with 11 C-methionine uptake in WHO grade II and III gliomas with IDH1 mutations. Brain Tumor Pathol 2017;34:20-7. [Crossref] [PubMed]
5. Jiao Y, Killela PJ, Reitman ZJ, Rasheed AB, Heaphy CM, de Wilde RF, et al. Frequent ATRX, CIC, FUBP1 and IDH1 mutations refine the classification of malignant gliomas. Oncotarget 2012;3:709-22. [Crossref] [PubMed]
6. Esteller M, Garcia-Foncillas J, Andion E, Goodman SN, Hidalgo OF, Vanaclocha V, Baylin SB, Herman JG. Inactivation of the DNA-repair gene MGMT and the clinical response of gliomas to alkylating agents. N Engl J Med 2000;343:1350-4. [Crossref] [PubMed]
7. Haque W, Teh C, Butler EB, Teh BS. Prognostic and predictive impact of MGMT promoter methylation status in high risk grade II glioma. J Neurooncol 2022;157:137-46. [Crossref] [PubMed]
8. Binabaj MM, Bahrami A. ShahidSales S, Joodi M, Joudi Mashhad M, Hassanian SM, Anvari K, Avan A. The prognostic value of MGMT promoter methylation in glioblastoma: A meta-analysis of clinical trials. J Cell Physiol 2018;233:378-86. [Crossref] [PubMed]
9. Arita H, Narita Y, Takami H, Fukushima S, Matsushita Y, Yoshida A, Miyakita Y, Ohno M, Shibui S, Ichimura K. TERT promoter mutations rather than methylation are the main mechanism for TERT upregulation in adult gliomas. Acta Neuropathol 2013;126:939-41. [Crossref] [PubMed]
10. Ozturk-Isik E, Cengiz S, Ozcan A, Yakicier C, Ersen Danyeli A, Pamir MN, Özduman K, Dincer A. Identification of IDH and TERTp mutation status using (1) H-MRS in 112 hemispheric diffuse gliomas. J Magn Reson Imaging 2020;51:1799-809. [Crossref] [PubMed]
11. Labussière M, Di Stefano AL, Gleize V, Boisselier B, Giry M, Mangesius S, Bruno A, Paterra R, Marie Y, Rahimian A, Finocchiaro G, Houlston RS, Hoang-Xuan K, Idbaih A, Delattre JY, Mokhtari K, Sanson M. TERT promoter mutations in gliomas, genetic associations and clinico-pathological correlations. Br J Cancer 2014;111:2024-32. [Crossref] [PubMed]
12. Yang X, Hu C, Xing Z, Lin Y, Su Y, Wang X, Cao D. Prediction of Ki-67 labeling index, ATRX mutation, and MGMT promoter methylation status in IDH-mutant astrocytoma by morphological MRI, SWI, DWI, and DSC-PWI. Eur Radiol 2023;33:7003-14. [Crossref] [PubMed]
13. Le Bihan D. The 'wet mind': water and functional neuroimaging. Phys Med Biol 2007;52:R57-90. [Crossref] [PubMed]
14. Tang L, Zhou XJ. Diffusion MRI of cancer: From low to high b-values. J Magn Reson Imaging 2019;49:23-40. [Crossref] [PubMed]
15. Abdalla G, Sanverdi E, Machado PM, Kwong JSW, Panovska-Griffiths J, Rojas-Garcia A, Yoneoka D, Yousry T, Bisdas S. Role of diffusional kurtosis imaging in grading of brain gliomas: a protocol for systematic review and meta-analysis. BMJ Open 2018;8:e025123. [Crossref] [PubMed]
16. Fan Z, Guo J, Zhang X, Chen Z, Wang B, Jiang Y, Li Y, Wang Y, Yang G, Wang X. Non-Gaussian diffusion metrics with whole-tumor histogram analysis for bladder cancer diagnosis: muscle invasion and histological grade. Insights Imaging 2024;15:138. [Crossref] [PubMed]
17. Le Bihan D. Intravoxel incoherent motion perfusion MR imaging: a wake-up call. Radiology 2008;249:748-52. [Crossref] [PubMed]
18. Jensen JH, Helpern JA, Ramani A, Lu H, Kaczynski K. Diffusional kurtosis imaging: the quantification of non-gaussian water diffusion by means of magnetic resonance imaging. Magn Reson Med 2005;53:1432-40. [Crossref] [PubMed]
19. Zhang K, Dai Y, Liu Y, Tao J, Pan Z, Xie L, Wang S. Soft tissue sarcoma: IVIM and DKI parameters correlate with Ki-67 labeling index on direct comparison of MRI and histopathological slices. Eur Radiol 2022;32:5659-68. [Crossref] [PubMed]
20. Jin YN, Zhang Y, Cheng JL, Zheng DD, Hu Y. Monoexponential, Biexponential, and stretched-exponential models using diffusion-weighted imaging: A quantitative differentiation of breast lesions at 3.0T. J Magn Reson Imaging 2019;50:1461-7. [Crossref] [PubMed]
21. Karaman MM, Sui Y, Wang H, Magin RL, Li Y, Zhou XJ. Differentiating low- and high-grade pediatric brain tumors using a continuous-time random-walk diffusion model at high b-values. Magn Reson Med 2016;76:1149-57. [Crossref] [PubMed]
22. Ingo C, Magin RL, Colon-Perez L, Triplett W, Mareci TH. On random walks and entropy in diffusion-weighted magnetic resonance imaging studies of neural tissue. Magn Reson Med 2014;71:617-27. [Crossref] [PubMed]
23. Zhou XJ, Gao Q, Abdullah O, Magin RL. Studies of anomalous diffusion in the human brain using fractional order calculus. Magn Reson Med 2010;63:562-9. [Crossref] [PubMed]
24. Karaman MM, Zhang J, Xie KL, Zhu W, Zhou XJ. Quartile histogram assessment of glioma malignancy using high b-value diffusion MRI with a continuous-time random-walk model. NMR Biomed 2021;34:e4485. [Crossref] [PubMed]
25. Mao C, Hu L, Jiang W, Qiu Y, Yang Z, Liu Y, Wang M, Wang D, Su Y, Lin J, Yan X, Cai Z, Zhang X, Shen J. Discrimination between human epidermal growth factor receptor 2 (HER2)-low-expressing and HER2-overexpressing breast cancers: a comparative study of four MRI diffusion models. Eur Radiol 2024;34:2546-59. [Crossref] [PubMed]
26. Sui Y, Wang H, Liu G, Damen FW, Wanamaker C, Li Y, Zhou XJ. Differentiation of Low- and High-Grade Pediatric Brain Tumors with High b-Value Diffusion-weighted MR Imaging and a Fractional Order Calculus Model. Radiology 2015;277:489-96. [Crossref] [PubMed]
27. Sui Y, Xiong Y, Jiang J, Karaman MM, Xie KL, Zhu W, Zhou XJ. Differentiation of Low- and High-Grade Gliomas Using High b-Value Diffusion Imaging with a Non-Gaussian Diffusion Model. AJNR Am J Neuroradiol 2016;37:1643-9. [Crossref] [PubMed]
28. Gao A, Zhang H, Yan X, Wang S, Chen Q, Gao E, Qi J, Bai J, Zhang Y, Cheng J. Whole-Tumor Histogram Analysis of Multiple Diffusion Metrics for Glioma Genotyping. Radiology 2022;302:652-61. [Crossref] [PubMed]
29. Ge X, Gan T, Yang Z, Liu G, Hu W, Ma W, Bai Y, Xiong Y, Li M, Zhao J, Zhou L, Li J, Li D, Wang X, Zhang J. Whole-tumor histogram analysis of synthetic magnetic resonance imaging predicts isocitrate dehydrogenase mutation status in gliomas. Quant Imaging Med Surg 2024;14:2225-39. [Crossref] [PubMed]
30. Zhang W, Zhou Y, Xu XQ, Kong LY, Xu H, Yu TF, Shi HB, Feng Q. A Whole-Tumor Histogram Analysis of Apparent Diffusion Coefficient Maps for Differentiating Thymic Carcinoma from Lymphoma. Korean J Radiol 2018;19:358-65. [Crossref] [PubMed]
31. Sun Y, Yang Z, Deng K, Geng Y, Hu X, Song Y, Jiang R. Histogram analysis of quantitative susceptibility mapping and apparent diffusion coefficient for identifying isocitrate dehydrogenase genotypes and tumor subtypes of adult-type diffuse gliomas. Quant Imaging Med Surg 2023;13:8681-93. [Crossref] [PubMed]
32. Yang Q, Reutens DC, Vegh V. Generalisation of continuous time random walk to anomalous diffusion MRI models with an age-related evaluation of human corpus callosum. Neuroimage 2022;250:118903. [Crossref] [PubMed]
33. Barrick TR, Spilling CA, Ingo C, Madigan J, Isaacs JD, Rich P, Jones TL, Magin RL, Hall MG, Howe FA. Quasi-diffusion magnetic resonance imaging (QDI): A fast, high b-value diffusion imaging technique. Neuroimage 2020;211:116606. [Crossref] [PubMed]
34. Wáng YXJ. An explanation for the triphasic dependency of apparent diffusion coefficient (ADC) on T2 relaxation time: the multiple T2 compartments model. Quant Imaging Med Surg 2025;15:3779-91. [Crossref] [PubMed]
35. Dang L, Yen K, Attar EC. IDH mutations in cancer and progress toward development of targeted therapeutics. Ann Oncol 2016;27:599-608. [Crossref] [PubMed]
36. Brat DJ, Verhaak RG, Aldape KD, Yung WK, Salama SR, et al. Comprehensive, Integrative Genomic Analysis of Diffuse Lower-Grade Gliomas. N Engl J Med 2015;372:2481-98. [Crossref] [PubMed]
37. Turkalp Z, Karamchandani J, Das S. IDH mutation in glioma: new insights and promises for the future. JAMA Neurol 2014;71:1319-25. [Crossref] [PubMed]
38. Cindil E, Sendur HN, Cerit MN, Erdogan N, Celebi F, Dag N, Celtikci E, Inan A, Oner Y, Tali T. Prediction of IDH Mutation Status in High-grade Gliomas Using DWI and High T1-weight DSC-MRI. Acad Radiol 2022;29:S52-62. [Crossref] [PubMed]
39. Zhang HW, Zhang HB, Liu XL, Deng HZ, Zhang YZ, Tang XM, Lin F, Huang B. Clinical Assessment of Magnetic Resonance Spectroscopy and Diffusion-Weighted Imaging in Diffuse Glioma: Insights Into Histological Grading and IDH Classification. Can Assoc Radiol J 2024;75:868-77. [Crossref] [PubMed]
40. Xie Y, Li S, Shen N, Gan T, Zhang S, Liu WV, Zhu W. Assessment of Isocitrate Dehydrogenase 1 Genotype and Cell Proliferation in Gliomas Using Multiple Diffusion Magnetic Resonance Imaging. Front Neurosci 2021;15:783361. [Crossref] [PubMed]
41. Weller M, Reifenberger G. Beyond the World Health Organization classification of central nervous system tumors 2016: what are the new developments for gliomas from a clinician's perspective? Curr Opin Neurol 2020;33:701-6. [Crossref] [PubMed]
42. Cheng Y, Song S, Wei Y, Xu G, An Y, Ma J, Yang H, Qi Z, Xiao X, Bai J, Xu L, Hu Z, Sun T, Wang L, Lu J, Lin Q. Glioma Imaging by O-(2-18F-Fluoroethyl)-L-Tyrosine PET and Diffusion-Weighted MRI and Correlation With Molecular Phenotypes, Validated by PET/MR-Guided Biopsies. Front Oncol 2021;11:743655. [Crossref] [PubMed]
43. Choi YS, Ahn SS, Kim DW, Chang JH, Kang SG, Kim EH, Kim SH, Rim TH, Lee SK. Incremental Prognostic Value of ADC Histogram Analysis over MGMT Promoter Methylation Status in Patients with Glioblastoma. Radiology 2016;281:175-84. [Crossref] [PubMed]
44. Romano A, Calabria LF, Tavanti F, Minniti G, Rossi-Espagnet MC, Coppola V, Pugliese S, Guida D, Francione G, Colonnese C, Fantozzi LM, Bozzao A. Apparent diffusion coefficient obtained by magnetic resonance imaging as a prognostic marker in glioblastomas: correlation with MGMT promoter methylation status. Eur Radiol 2013;23:513-20. [Crossref] [PubMed]
45. Jiang S, Rui Q, Wang Y, Heo HY, Zou T, Yu H, Zhang Y, Wang X, Du Y, Wen X, Chen F, Wang J, Eberhart CG, Zhou J, Wen Z. Discriminating MGMT promoter methylation status in patients with glioblastoma employing amide proton transfer-weighted MRI metrics. Eur Radiol 2018;28:2115-23. [Crossref] [PubMed]
46. Nakamura M, Masutomi K, Kyo S, Hashimoto M, Maida Y, Kanaya T, Tanaka M, Hahn WC, Inoue M. Efficient inhibition of human telomerase reverse transcriptase expression by RNA interference sensitizes cancer cells to ionizing radiation and chemotherapy. Hum Gene Ther 2005;16:859-68. [Crossref] [PubMed]
47. Rohwer N, Zasada C, Kempa S, Cramer T. The growing complexity of HIF-1α's role in tumorigenesis: DNA repair and beyond. Oncogene 2013;32:3569-76. [Crossref] [PubMed]
48. Killela PJ, Reitman ZJ, Jiao Y, Bettegowda C, Agrawal N, Diaz LA Jr, et al. TERT promoter mutations occur frequently in gliomas and a subset of tumors derived from cells with low rates of self-renewal. Proc Natl Acad Sci U S A 2013;110:6021-6. [Crossref] [PubMed]
49. Low KC, Tergaonkar V. Telomerase: central regulator of all of the hallmarks of cancer. Trends Biochem Sci 2013;38:426-34. [Crossref] [PubMed]
50. Yamashita K, Hatae R, Hiwatashi A, Togao O, Kikuchi K, Momosaka D, Yamashita Y, Kuga D, Hata N, Yoshimoto K, Suzuki SO, Iwaki T, Iihara K, Honda H. Predicting TERT promoter mutation using MR images in patients with wild-type IDH1 glioblastoma. Diagn Interv Imaging 2019;100:411-9. [Crossref] [PubMed]
51. Park YW, Ahn SS, Park CJ, Han K, Kim EH, Kang SG, Chang JH, Kim SH, Lee SK. Diffusion and perfusion MRI may predict EGFR amplification and the TERT promoter mutation status of IDH-wildtype lower-grade gliomas. Eur Radiol 2020;30:6475-84. [Crossref] [PubMed]