Introduction

Gliomas are the most frequent primary brain tumors in adults with an overall age-adjusted population incidence rate of 4.67–5.73 per 100,000, without remarkable change over the past two decades [1, 2]. Adult diffuse gliomas are graded as CNS WHO grades 2, 3, or 4 according to the degree of malignancy; the higher the grade, the more malignant, and the poorer the prognosis. The primary treatment for IDH-mutant grade 2 and 3 diffuse gliomas is maximal safe resection followed by active surveillance, or radiation followed by procarbazine, lomustine, and vincristine, while radiation with adjuvant temozolomide is often used in the treatment of IDH-mutant astrocytoma, grade 4 and IDH-wildtype glioblastomas (grade 4) [3]. Despite the development of therapeutics, the prognosis for grade 4 gliomas remains unfavorable. The reported median survival of grades 2, 3, and 4 gliomas are approximately 5–17 years, 2–8 years, and <1 year, respectively [1].

Recent advances in molecular genetics have allowed for a more concise definition of CNS tumors based on molecular genetic features. The diagnostic criteria for adult diffuse gliomas now rely on the presence or absence of an isocitrate dehydrogenase (IDH) gene mutation and 1p/19q codeletion, which were adopted in the 2016 edition of WHO classification [4], with the addition of several updates in the latest WHO classification in 2021 (WHO CNS5) [5]. For example, a diagnosis of IDH-wildtype grade 4 diffuse glioma (glioblastoma) includes 3 new molecular genetic criteria: concurrent gain of the whole chromosome 7 and loss of the whole chromosome 10 (+7/−10), telomerase reverse transcriptase (TERT) promoter mutation, and epidermal growth factor receptor (EGFR) amplification. Meanwhile, oligodendrogliomas are classified only into grade 2 or 3 [6]. Due to its rarity, an IDH-wildtype lower-grade (grade 2 or 3) astrocytoma is no longer regarded as a distinct tumor type in WHO CNS5 [6].

Due to these recent revisions and reclassifications of WHO criteria, a diagnosis of grade 4 adult diffuse gliomas should be made even if there is no histological evidence of necrosis or microvascular proliferation. The presence of grade 4 diffuse glioma among tumors diagnosed as lower-grade glioma in previous criteria evokes a challenging problem in radiological diagnosis. Traditionally, a rim-enhancing mass with central necrosis has been considered a typical appearance for grade 4 diffuse glioma, including IDH-wildtype glioblastomas and grade 4 IDH-mutant astrocytomas [7, 8]. However, even tumors without imaging evidence of necrosis could be diagnosed as grade 4 diffuse gliomas according to WHO CNS5 [5].

Despite the difficulty of imaging diagnosis, it is meaningful to diagnose these “molecular” grade 4 diffuse gliomas because these updates are based on advances in molecular genetic findings that revealed histologically lower-grade gliomas with these molecular abnormalities have a prognosis similar to those of histological grade 4 diffuse gliomas [9,10,11]. Moreover, these molecular genetic features are potential targets for molecularly targeted therapies currently under development, although the results have not yet been satisfactory [12, 13]. Therefore, imaging biomarkers on preoperative MRI for predicting molecular grade 4 diffuse glioma have been warranted to formulate a treatment strategy. Significant efforts have been made to differentiate grade 4 and lower-grade diffuse gliomas, and DWI has provided useful imaging biomarkers including ADC [1 and Fig. 2, respectively.

Table 4 Multivariate logistic regression analysis
Full size table
Fig. 1
figure 1

An 84-year-old man with IDH-wildtype histologically lower-grade astrocytic glioma with molecular evidence of CNS WHO grade 4 (concurrent gain of the whole chromosome 7 and loss of the whole chromosome 10 and TERT promoter mutation). The tumor shows hyperintensity on the fluid-attenuated inversion recovery image (a), and the median and entropy of the normalized apparent diffusion coefficient are 1.28 and 6.19, respectively (b). No evidence of necrosis or contrast enhancement is found on the subtraction T1-weighted image (c)

Full size image
Fig. 2
figure 2

A 26-year-old woman with IDH-mutant histologically lower-grade astrocytoma without homozygous deletions of CDKN2A/B. The tumor shows hyperintensity on the fluid-attenuated inversion recovery image (a), and the median and entropy of the normalized apparent diffusion coefficient are 2.37 and 6.89, respectively (b). No evidence of necrosis or contrast enhancement is found on the subtraction T1-weighted image (c)

Full size image

Subclass analyses between the Mol-4 and Mol-2/3 astrocytomas

The demographic and pathological findings and the results of ADC histogram analyses are summarized in Supplementary Tables 1 and 2, respectively. Age at diagnosis was significantly higher, and the rate of IDH-mutant tumors was significantly lower in the Mol-4 astrocytomas compared with the Mol-2/3 counterparts (median 64.5 [range, 37–84] years vs. 33 [18–60] years, p < 0.001; IDH-mutant vs. IDH-wildtype: 1:17 vs. 24:3, p < 0.001). The age at diagnosis and median of ADC were selected for the stepwise multivariate logistic regression analysis, and the AUC of the model was 0.98 (0.94–1) (Supplementary Table 3). The mean precision, recall, and F1 score using LOOCV and 1000-bootstrap** (mean (95%CI)) were 0.86 (0.86–0.87), 0.85 (0.85–0.86), and 0.86 (0.85–0.86), respectively. Although the median of ADC tended to be lower in the Mol-4 astrocytomas, the p-value did not reach statistical significance.

Discussion

In this study, we compared ADC histogram profiles in histologically lower-grade adult diffuse gliomas between the presence (the Mol-4 group) and absence (the Mol-2/3 group) of molecular evidence for CNS WHO grade 4 in accordance with WHO CNS5. Many ADC histogram profiles were significantly different between the two groups, and the entropy of normalized ADC and age at diagnoses were found as independent predictive factors for the Mol-4 group with an AUC of 0.92. On the other hand, while several ADC histogram profiles were significantly different between the groups, only age at diagnosis was the independent predictive factor for the Mol-4 astrocytoma group in the comparison between the Mol-4 and Mol-2/3 astrocytomas.

In the present study, we found that 1st-order ADC histogram profiles (i.e., maximum, 90th percentile, median, mean, and 10th percentile) were significantly lower in the Mol-4 group than in the Mol-2/3 group. These results were similar to the results of the subclass analyses of astrocytomas. Traditionally, a lower ADC has been considered to be associated with increased cellularity [26]. Furthermore, many studies have reported that lower ADC is a negative prognostic biomarker in gliomas independent of the grade. In the meta-analysis of studies published between 2006 and 2010, Zulfigar et al. [27] demonstrated that a low minimum ADC was associated with a shorter survival rate in patients with grades III and IV diffuse glioma when cases were stratified according to the grades. Cuccarini et al. [28] reported that a low minimum ADC was associated with a shorter overall survival in patients with grades II and III diffuse glioma when cases were stratified according to the grades. Hilario et al. [29] found that older age and a low median ADC were associated with shorter overall survival independent of tumor grade in their study including patients with grades II, III, and IV diffuse gliomas. Despite the consistency of the results, it should be noted that diffuse gliomas in these studies were diagnosed according to the previous WHO classifications without the latest molecular updates; thus, histologically lower-grade molecular grade 4 diffuse gliomas (i.e., the Mol-4 group) were likely included in their grades II or III diffuse glioma groups [26,27,28,29]. Therefore, the variations in ADC and prognosis among the “same” grade diffuse gliomas in these studies may have been affected, at least in part, by the differences in the presence or absence of the molecular features of grade 4, as shown in the present study.

The multivariate stepwise logistic regression analysis revealed that age at diagnosis and the entropy of normalized ADC were independent parameters for predicting the Mol-4 group. The second-order parameters derived from the whole tumor ADC histograms, namely kurtosis, skewness, and entropy, have been evaluated for tumor grading and for estimating the proliferative potential of the tumors [30,31,32]. Kurtosis represents the distribution peakedness of the intensity for the parameter within the tissue. Skewness represents the distribution asymmetry. Entropy represents the intensity predictability of the parameter and reflects textural variation. Differences in the entropy of normalized ADC in the present study may have reflected the difference in microstructural heterogeneity between the two groups. Although the p-value of the entropy of ADC did not reach statistical significance and only age at diagnosis was an independent predictive factor for the Mol-4 astrocytoma group in the subclass analysis without oligodendrogliomas, the entropy of ADC was also selected in the stepwise selection. The higher patient age in the Mol-4 group compared with the Mol-2/3 group was in line with the study by Reuss et al. [33], although they did not compare the patient age statistically. Further studies with a larger number of cases are warranted to verify these results.

While ADC histogram profiles were shown to be promising imaging biomarkers in differentiating Mol-4 group from Mol-2/3 group, we could not verify the difference in the frequency of the gyriform infiltration on FLAIR imaging between the two groups. The gyriform infiltration was reported to be present in 16/31 patients (51.6%) in the Mol-4 group and 0/151 patients (0%) in the Mol-2/3 group with a statistically significant difference in the frequency by a French group [18]. The discrepancy between the results of their study and the present study should be further investigated by another group since other than one study by the same French group with the patients from the same institution [34] and the present study, no study has examined the frequency of the gyriform infiltration in the Mol-4 group. It should be noted that although all of their Mol-4 group tumors corresponded to lower-grade IDH-wildtype astrocytomas in the previous WHO 2016 classification [18], the high frequency of gyriform infiltration in lower-grade IDH-wildtype astrocytomas has not been reported from other groups.

We recognize several limitations in the present study. First, this was a single institutional retrospective study, which limited the number of patients. Second, MRI examinations were performed using multiple scanners at different field strengths. However, to minimize the influence of the machine differences, we normalized ADC parameters. Further, ADC is theoretically insensitive to the difference in field strength, even though a higher field strength may benefit from a higher contrast-to-noise ratio [35]. Third, most of the Mol-4 group consisted of IDH-wildtype diffuse gliomas in the present study. However, this reflects the real-world frequency because we included consecutive patients, and the similar frequency of IDH-wildtype diffuse gliomas has been reported in other studies [18, 33]. Finally, we could not evaluate the correlation between histogram profiles and patient survival. The association with prognosis needs to be investigated in future studies.

Conclusion

Whole tumor histogram-derived ADC profiles may be promising imaging biomarkers on preoperative brain MRI with significant differences between the presence and absence of molecular features for grade 4 in histologically lower-grade adult diffuse gliomas.