In this retrospective single-center study, consecutive patients aged 18 years or older who underwent an MRI evaluation in relation to endometriosis performed at our hospital between February 2017 and June 2022 were identified. Exclusion criteria were prior hysterectomy and incomplete patient data, i.e. externally referred patient with no precise information about symptoms or medical history. The study was approved by the local ethics committee (Cantonal Ethics Committee Zurich) and informed consent was waived.

MRI scans were performed on 3.0 or 1.5 Tesla MR scanners (Siemens Skyra, Sola or Vida fit, Siemens Healthineers, Erlangen, Germany, and GE Medical Systems Discovery MR750w, GE Medical Systems, Milwaukee, WI, USA) using a dedicated pelvic MRI protocol in accordance with current guidelines [10, 11], encompassing high-resolution 2D T2-weighted TSE sequences in three orientations, 3D T1-weighted GRE sequences with and without fat-suppression and 3D T1-weighted GRE sequences with fat-suppression after contrast agent administration. In most cases, an IV anti-peristaltic agent (butylscopolamin 20 mg/ml) was injected to reduce artifacts due to bowel peristaltics before the examination. No rectal or vaginal opacification was performed.

Before the study image analysis, a training session on cases that were not part of the cohort was conducted. Subsequently, two radiologists with 5 and 3 years of experience in pelvic MRI independently reviewed the MRI images on a Picture Archiving and Communication System (PACS) workstation, while being blinded to clinical or histopathological information, except for the fact that the MRI was performed for the purpose of evaluating endometriosis.

The localization and severity grading of the endometriotic lesions and adhesions were assessed using the criteria outlined in the #Enzian classification in the publication by Keckstein et al. [9]:

For the peritoneum (P), superficial peritoneal implants with a sub-peritoneal invasion of less than 5 mm are considered. They are categorized based on the sum of all maximal diameters as: P1 < 3 cm; P2 = 3–7 cm; P3 > 7 cm.

Regarding the ovaries (O), all endometriomas and infiltrating ovarian surface foci with a size of 5 mm or larger are assessed. They are categorized based on the sum of all maximal diameters as: O1 < 3 cm; O2 = 3–7 cm; O3 > 7 cm.

The evaluation of the tubo-ovarian condition (T) involves the presence of adhesions between the ovary and pelvic sidewall with or without tubo-ovarian adhesions. The classification includes the following categories: T1 for adhesions between the ovary and pelvic sidewall, T1 plus adhesions to the uterus or isolated adhesions between the adnexa (ovaries and fallopian tubes) and uterus (T2), and T2 plus adhesions to the uterosacral ligaments (USLs) and/or bowel, or isolated adhesions between the adnexa and USLs and/or bowel (T3).

Deep endometriosis (DE) refers to implants with sub-peritoneal infiltration greater than 5 mm. The #Enzian score classifies these lesions based on their site and the involved organs. They are categorized into three compartments: Compartment A, which includes the vagina, recto-vaginal space, or retrocervical area (measured in the sagittal plane); Compartment B, encompassing the uterosacral and cardinal ligaments or pelvic sidewall (measured in the axial plane); and Compartment C, comprising the rectal wall up to 16 cm from the anal verge (measured in the sagittal plane). Lesions within each compartment are further described based on the sum of all maximal diameters: A/B/C1 < 1 cm, A/B/C2 = 1–3 cm, and A/B/C3 > 3 cm. The description of each compartment is provided separately.

For adenomyosis and other extragenital deep endometriosis, the #Enzian score includes additional categories: FA for uterine adenomyosis, defined as thickening of the myometrium-endometrium junction line greater than 12 mm; FB for bladder lesions involving the muscular layer; FU for ureteral lesions involving the muscular layer (both extrinsic and intrinsic); FI for lesions in the sigmoid colon, coecum, or ileum located above 16 cm from the anus; and F(…) for other lesions, such as those on the diaphragms, liver, or abdominal wall.

Paired organ compartments, including compartment O, T and B, were assessed, and documented separately for each side (left/right).

The #Enzian classification also encompasses additional information about the mobility of the ovaries and tubes, as well as tubal patency. However, these aspects are not evaluable on MRI, and were therefore not included in the present study. No differentiation between intrinsic or extrinsic ureteral endometriosis was made due to the small number of cases in the study population.

Additionally, as recommended by Manganaro et al. [12], the presence (SA 1) or absence (SA 0) of a sactosalpinx was evaluated separately for each side. Moreover the more experienced radiologist of the two readers noted the type of adenomyosis, following the readily applicable criteria proposed by Bazot et al. [13], who succinctly defined three types: internal adenomyosis (FA(i)), adenomyomas (FA(a)), and external adenomyosis (FA(e)). However, no further subclassifications of these types of adenomyosis were made to ensure a sufficient number of cases for each type. No correlation of the MRI findings for adenomyosis to histopathology could be established, as the specific type of adenomyosis is not routinely evaluated histopathologically.

Surgical interventions were performed by gynecologic surgeons with extensive experience in endometriosis surgery. Since MRI assessments are routinely conducted to plan the surgical intervention, the surgeons were aware of the preoperative evaluation findings. Removed endometriotic lesions were subsequently confirmed through histological examination. In cases where no endometriosis was observed during surgery, there was no assignment of a surgical #Enzian score since the condition was not encountered.

The dataset was subjected to descriptive statistics for analysis, using the statistical software R.17 [14] for data analysis. Level of significance was set to 5%.

To assess inter-reader agreement on the #Enzian MRI classification between the two radiologists Cohen’s kappa coefficients (κ) with 95% confidence intervals were computed. Agreement values falling between 0.81 and 1.00 were considered to indicate excellent (or ‘almost perfect’) agreement, 0.61–0.80 indicated substantial agreement, 0.41–0.60 represented moderate agreement, 0.21–0.40 denoted fair agreement, and 0.01–0.20 signified slight agreement [15]. Ratings from 0 to 3 in the compartments P, O, T, A, B and C and the dichotomization with values of 0 indicating the absence and 1 indicating the presence of a sactosalpinx (SA) were used to analyze the inter-reader agreement. For the compartment FA ratings were treated as binary, without taking size into consideration. The compartment-specific inter-reader agreement analysis excluded #Enzian locations FB, FI, FU, and F(…) due to the low number of MR-positive findings. Inter-group comparisons between the different types of adenomyosis were performed using Fisher’s exact test excluding patients with two different types of adenomyosis.

The surgical #Enzian classification served as the gold standard for evaluating the preoperative MRI #Enzian classification. For each compartment, measures such as accuracy (ACC), agreement (Cohen’s kappa coefficient, κ), sensitivity (SENS), specificity (SPEC), positive predictive value (PPV), and negative predictive value (NPV) were computed. This analysis only evaluated patients from the study population who had complete surgical #Enzian scores, implemented since January 2021. Due to the relatively small number of patients in the cohort subset, all ratings were treated as binary. Moreover, the analysis excluded #Enzian locations FB, FI, FU and F(…) due to the low number of MR-positive findings.