The pelvic MRI characteristics in patients of different morphology groups are summarised in Table 3. Haematometra was more common in patients with type I CCM than in those with type II CCM (p < 0.001). Moreover, the mean sagittal length was longer in patients with type I CCM than in those with type II CCM (p < 0.001). No significant differences were observed between the different groups with respect to other MRI features such as uterine volume, haematosalpinx, and degree of vaginal dysgenesis (p > 0.05). The detailed MRI findings of the entire patient population are described below.

Morphologically, the cervixes were categorised into three general types. Type I denoted a cervix with cervical canals (42 patients; 53%) and was further subdivided into internal ostial atresia (Fig. 3a–c) and external ostial atresia (Fig. 3d, e). Type II (18 patients; 23%) was marked by a cervix with invisible cervical canals, displaying homogeneous solid isointensity compared to the outer myometrium [23] on the same image slice (Fig. 4a, b), or displaying mixed signal (Fig. 4c, d). Type III (19 patients; 24%) was defined by a blind end in the lower part of the uterine corpus. Some patients exhibited a normal uterine morphology (Fig. 5a), while others showed abnormalities marked by numerous haemorrhage signals in the dilated uterine cavity (Fig. 5b). Notably, patients with type I tended to have longer cervixes than those with type II (5.6 ± 3.2 vs. 1.9 ± 0.6 cm, p < 0.001).

Magnetic resonance images of different subgroups of patients with type I congenital cervical malformations. Sagittal T2-weighted (a) and axial T2-weighted (b) and T1-weighted (c) images acquired from a 12-year-old patient with external ostial atresia. Sagittal and axial T2-weighted images reveal an enlarged cervix (white arrows) with a multi-layered differentiation signal, demonstrating haematocele presence, while T1-weighted images exhibit a prominent hyperintense signal in the cervix (white arrows). Sagittal T2-weighted (d) and axial T2-weighted (e) images acquired from a 13-year-old patient with internal ostial atresia. Sagittal and axial T2-weighted images show internal ostial atresia (white arrows), while the cervical canal exhibits a multi-layered differentiation signal (red arrows)

Magnetic resonance images depicting four individuals classified as having type II congenital cervical malformations. The sagittal T2-weighted images of these patients revealed the absence of luminal structures. Among these, two displayed homogeneous signals (red arrowheads in a and b), while the remaining two exhibited conspicuous mixed signals (white arrowheads in c and d)

Magnetic resonance images illustrating two subgroups of patient with type III congenital cervical malformations. The images clearly depict the normal body of the uterus (white arrowheads in a and b) along with the absence of the cervix (white arrows in a and b). Some patients exhibited an evident abnormal uterine corpus characterised by numerous haemorrhage signals (white arrowheads in c and d) as well as a missing cervix (white arrows in c and d)

The signal-based features of the cervix were generally classified into three types: no signal (19 patients; 24%) (Fig. 5), no-evident layer differentiation (18 patients; 23%) (Fig. 4), and evident layer differentiation with or without a haematocele signal inside the cervical canal (42 patients; 53%) (Fig. 3). Clear layer differentiation within the cervix was common among patients with type I, and was often accompanied by a haematocele signal inside the cervical canals. Within this group, 45.2% of patients displayed three-layer differentiation, while 54.8% exhibited four-layer differentiation. On the other hand, patients with type II typically lacked evident layer differentiation within the cervix, and those with type III showed no signal owing to the blind end in the lower part of the uterine corpus.

Our patients were also classified according to Rock et al’s system based on MRI findings. Nineteen patients had cervical aplasia, while 41 had cervical obstruction; moreover, 15 had a fibrous cord. However, it was challenging to apply Rock et al’s classification to four of the patients. One patient did not exhibit cervical aplasia or obstruction but merely had an internal ostium atresia (PUMCH classification: type Ib). The remaining three each had a small cervix that did not align with the classification criteria (PUMCH classification: type II, they do have a cervix but invisible cervical canals on MR images). None of the patients were diagnosed with cervical fragmentation.

Uterine malformations were detected in 16 patients; these included uterine agenesis (two; 12.5%), bicornuate uterus (one; 6.2%), uterus duplex (three; 18.8%), complete septate uterus (two; 12.5%), incomplete septate uterus (two; 12.5%), and rudimentary horn uterus (six; 37.5%). Haematometra, characterised by bleeding within the uterine cavity accompanied by dilatation thereof, was observed in 67 patients. Among those, haemorrhaging was observed in 41 patients with type I CCM, 10 with type II, and 16 with type III (p < 0.001). The median uterine volume among patients with type I CCMs was 63.3 ± 40.7 mL, while the median volumes in patients with types II and III were 40.8 ± 15.4 and 85.0 ± 56.9 mL, respectively. However, these differences were not statistically significant (p = 0.050). Additionally, adenomyosis was present in three of the patients in this study, all of whom were type II.

The maximum cross-sectional areas of the ovaries of patients with types I, II, and III CCM were 4.0 ± 2.1, 5.7 ± 4.3, and 4.4 ± 1.9 cm², respectively (p = 0.095). Ovarian endometriosis was diagnosed in 20 patients, with bilateral involvement in 10, right ovary only in four, and left ovary only in six. Additionally, the incidence of endometriosis did not differ significantly among patients with different types of CCM (p = 0.147). Pelvic endometriosis was detected in one patient, predominantly located in the posterior wall of the uterus, uterorectal fossa, fundal ligament, left ovary, and fallopian tube. Haematosalpinx was identified in 29 patients, including 15 with type I, six with type II, and eight with type III (p = 0.842); this condition typically presented as a tortuous and dilated fallopian tube that showed high signal intensity on T1WI and variable signal intensity on T2WI (i.e., either high or low).

The distal vagina was discernible on MRI in only 28 of the 79 patients. Most patients (74; 94%) also had complete vaginal atresia, including 38 with type I, 17 with type II, and 19 with type III (p = 0.661). One patient had a normal vagina.

Six patients exhibited congenital spinal deformity, while another five presented with congenital renal malformations that included conditions such as renal agenesis and duplication. Additionally, six patients had a documented history of congenital anal atresia.

The surgical methods undergone by the patients were classified into two groups. The first was the hysterectomy group that included 30 patients; six had type I CCM, 13 had type II, and 11 had type III. Of the six patients with type I, two had uterine malformations and four opted for total hysterectomy owing to thick or poor interstitial tissue. In contrast, patients with types II and III predominantly underwent hysterectomy (p < 0.001), primarily owing to anatomical factors such as the greater distance from the vaginal acupoint to the uterine corpus or the extremely thin space between the rectum and bladder. These anatomical considerations likely rendered other treatment options less feasible or effective. The other surgical group comprised 49 patients who underwent procedures aimed at restoring normal uterovaginal anatomical structure, including uterovaginal anastomosis or cervicoplasty. This group comprised six patients with type I CCM, five with type II, and eight with type III. Furthermore, pelvic adhesions were discovered in 57 patients (27, 15, and 15 with types I, II, and III, respectively; p = 0.266).