The characteristics of 21 AE patients and 36 medulloblastoma patients are presented in Table 1. Patients with AE were younger than those with medulloblastoma (5.1 ± 4.7 [0–16] vs. 6.9 ± 4.2 [1,2,3,4,5,6,7,8,9,10,11,12,13,14,15,16,17,18] years, p = 0.056). A significantly greater proportion of the medulloblastoma patients received adjuvant chemotherapy (p < 0.001), whereas the radiotherapy rates did not differ between the two groups (p = 0.14). No significant differences in sex or surgical treatment of the primary tumor were found between the groups.

During follow-up, the incidence of disease progression was significantly greater in AE patients than in medulloblastoma patients (16/21 [76.2%] vs. 10/36 [27.8%], p < 0.001). In terms of recurrence patterns, AE patients were more likely to experience local recurrence than medulloblastoma patients (12/16 [75.0%] vs. 3/10 [30.0%], p = 0.043). Among patients with progressive disease, the time from surgery to tumor recurrence/progression was significantly shorter in patients with AE (9.6 ± 8.1 [1,2,3,4,5,6,7,8,9,10,11,12,13,14,15,16,17,18,19,20,21,22,23,24,25,26] months) than in patients with medulloblastoma (35.5 ± 41.3 [7–137] months) (p = 0.007).

Four AE patients and three medulloblastoma patients were lost to long-term follow-up. The overall survival time was shorter for patients with AE than for those with medulloblastoma (82.8 ± 19.7 [6-188] vs. 145.2 ± 17.6 [13–225] months, p = 0.036).

Table 2 outlines the MRI features of 21 AEs and 36 medulloblastomas. During the presurgical MRI examination, AEs were larger in size than the medulloblastomas (56.1 ± 11.3 [39–83] vs. 48.1 ± 9.5 [27–77] mm, p = 0.007). No differences in signal intensity on T1WI, T2WI, or contrast-enhanced T1WI were found. However, AEs presented a more lobulated and ring-like enhancement pattern, whereas medulloblastomas presented a patchy and diffuse pattern (p < 0.001) (Figs. 3c, 4c and 5, and 6).

DWI was conducted in 18 AE patients and 35 medulloblastoma patients. The raw data for calculating the ADC were available for 17 AE patients and 33 medulloblastoma patients. The ADCmin values were significantly greater for AEs than for medulloblastomas (800.61 ± 173.34 [438.7–1092.3] vs. 572.82 ± 75.94 [329.3–715.7] ×10− 6 mm2/s, p < 0.001). Additionally, compared with medulloblastomas, AEs had a significantly greater ADC ratio (1.10 ± 0.27 [0.63–1.59] vs. 0.76 ± 0.11 [0.52–0.96], p < 0.001) and lower DWI ratio (1.24 ± 0.30 [0.71–1.99] vs. 1.53 ± 0.19 [1.23–1.96], p < 0.001). In the ROC curve analysis, the ADCmin cutoff for distinguishing AEs from medulloblastomas was 647.84 × 10− 6 mm2/s. ADCmin values above the cutoff indicated AEs, with 88.2% sensitivity, 87.9% specificity, 88.0% accuracy, and an AUC of 0.906. The ADC ratio had a cutoff of 0.980; ADC ratios above the cutoff were indicative of AEs, with 70.6% sensitivity, 100.0% specificity, 90.0% accuracy, and an AUC of 0.884. The DWI ratio had a cutoff of 1.275; values below the cutoff were indicative of AE, with 72.2% sensitivity, 97.1% specificity, 88.7% accuracy, and an AUC of 0.845. (Supplementary Table S1) No differences were observed in the ADC or DWI values of normal white matter between the two groups.

In terms of intratumoral morphology, the prevalence of the tumor central vein sign was significantly greater in patients with medulloblastomas than in those with AEs (15/36 [41.7%] vs. 1/21 [4.8%], p = 0.002) (Figs. 4a-c, 5 and 6). No significant difference was observed in the occurrence of intratumoral cysts/necrosis (p = 1.00) or hemorrhage (p = 0.33) between the two groups.

For peritumoral involvement, the prevalence of caudal extension through the foramen of Magendie on sagittal T2WI was greater for AEs than for medulloblastomas (21/21 [100.0%] vs. 19/36 [52.8%], p < 0.001), and the caudal tumor extension below the McRae line was longer in AEs than in medulloblastomas (14.36 ± 8.74 [4.0–33.0] vs. 7.90 ± 4.10 [3.5–19.0] mm, p = 0.016). In addition, the length/width ratio (base) (1.03 ± 0.84 [0.29–4.13] vs. 0.55 ± 0.21 [0.26-1.00], p = 0.005) and length/width ratio (mid) (2.39 ± 2.26 [0.56–10.67] vs. 0.90 ± 0.34 [0.42–1.40], p < 0.001) were significantly greater in AEs than in medulloblastomas.

Nearly all patients in both groups demonstrated lateral tumor extension into the foramen of Luschka (p = 1.00), except one patient with medulloblastoma. AEs more frequently encircled the brainstem than medulloblastomas did (14/21 [66.7%] vs. 2/36 [5.6%], p < 0.001) (Figs. 3a and b, 4b and 6). With respect to the involvement of peritumoral vessels, AEs exhibited a greater incidence of abutting or encasing the basilar artery (11/21 [52.4%] vs. 0/36 [0.0%], p < 0.001) (Fig. 3b), vertebral artery (18/21 [85.7%] vs. 8/36 [22.2%], p < 0.001) (Figs. 3a and 6), and PICA (18/21 [85.7%] vs. 10/36 [27.8%], p < 0.001) (Fig. 3a) than medulloblastomas did. No significant difference was identified in brainstem or middle cerebellar peduncle invasion (p = 0.37 and 0.078, respectively). Medulloblastomas tended to have more peritumoral brain edema than AEs did, although this difference was not statistically significant (p = 0.084).

With respect to distant involvement, leptomeningeal seeding at diagnosis was more common in patients with medulloblastomas than in those with AEs (8/36 [22.2%] vs. 0/20 [0.0%], p = 0.021). There was no difference in the incidence of hydrocephalus between the two groups (p = 0.33).