This single-centre retrospective study was approved by the Institutional Review Board of West China Second University Hospital (No. 2020174), and we pledged to abide by the declaration of Helsinki (2013 EDITION) in accordance with the relevant medical research rules of China in the study. Written informed consent regarding knowledge of adverse reactions to Gadolinium contrast was obtained from all patients prior to MR examination. All patient sensitive information was treated with strict secrecy and used solely for the purposes of this study.
From May 2014 to April 2019, a total of 5118 consecutive women with clinical and ultrasound findings indicating a suspicion of pregnancy-related diseases (RPOC [309/5118], CSP [3800/5118], and GTD [1009/5118]) were reviewed from our institutional database. A total of 3983 patients were excluded because they had unavailable DSA data for diagnosis and/or therapy, were treated by transcervical hysteroscopic resection, D&C, an intravenous injection of methotrexate, or conservative treatment, or were not treated in our hospital. In addition, 1020 patients were excluded because MRI was not performed before DSA. Thirty-six patients were excluded because they had non-pregnancy-related diseases identified by pathology (haematoma or blood clot [12/36], myoma degeneration [5/36], infections [19/36]). Thus, 79 patients were finally included in the data analysis according to the following inclusion criteria (Figure 1): (i) available MRI data acquired before DSA; (ii) available DSA data for diagnosis and/or therapy; and (iii) pregnancy-related diseases such as RPOC, GTD and CSP confirmed by pathology.
MRI examination was performed for patients with RPOC, CSP, and GTD before they were managed with D&C, transcervical hysteroscopic resection or laparoscopic resection. All MRI examinations were performed using 1.5T MRI systems (Philips Achieva 1.5T with Nova Dual HP, Best, the Netherlands) using an 8-channel phased-array body coil. The MRI protocol included axial T1-weighted fast spin echo imaging, coronal and sagittal T2-weighted fast spin echo imaging, axial T2-weighted fast spin echo imaging with fat suppression and diffusion-weighted imaging (DWI). Contrast-enhanced MRI was performed for 22 patients using T1-weighted imaging (T1WI) with fat saturation was performed in the axial, sagittal and coronal planes after the intravenous administration of 0.2 ml/kg body weight gadopentetate dimeglumine (Gd-DTPA, Magnevist; Bayer Schering, Berlin, Germany). The scanning parameters were summarized in Table 1.
Angiography and uterine artery embolization (UAE) were performed for 79 patients in our study using a therapeutic angiographic unit equipped with a digital flat panel detector system (Philips FD20 DSA) and nonionic contrast medium (iopamidol injection, 370 mg iodine per mL, Bracco Sine). All procedures were performed under conscious sedation (using midazolam and fentanyl) and local anaesthesia after a preprocedural evaluation performed by an anaesthesiologist. Moderate sedation (phenobarbital, 100 mg) and local anaesthesia were achieved with an intramuscular injection, and a 5-F vascular sheath (Radifocus; Terumo, Tokyo, Japan) was inserted into the right femoral artery using the Seldinger technique. Then, the uterine arteries were identified with DSA, and catheterization was performed selectively with a 5-F catheter (Radifocus; Terumo, Tokyo, Japan) along with an injection of 9 mL contrast material at a flow rate of 3 mL/sec. Angiograms were obtained through the catheter, and special attention was paid to the draining or feeding branches. The material used for embolization in this study was either a gelatine sponge (560-710 µm, Alicon) or polyvinyl alcohol (500-710 microns, 1CC, COOK Incorporated), or both. The embolization was complete when the maximal reduction in flow or stasis was visualized within the system. A final aortogram was obtained to assess the size of any residual vascular malformations with an injection of 6-12 mL contrast material at a flow rate of 1 mL/sec.
For MRI, in our study, the imaging sequences included T1WI, T2WI, T2WI-fat suppression, DWI and contrast-enhanced series. The flow void sign was difficult to observe on the T2WI-fat suppression, DWI and contrast-enhanced series due to hypointensity in the parametrium as well as the flow void. Therefore, we mainly analysed the flow-void sign on T1 and T2 turbo spin echo (TSE) images without fat suppression or contrast medium.
As previously described [24,28,29,30,31], the flow-void sign was defined as multiple dotlike, tubular structures or multiple serpentine with low signal intensity located within the myometrium and parametrium on both T1WI and T2WI in our study.
On one hand, two experienced gynaecological radiologists (reader 1, with 8 years; and reader 2, with 12 years) who were not aware of the DSA results reviewed all MRI images in consensus to evaluate the following traits for each pregnancy-related disease: (a) the morphological characteristics of the lesion (including location, signal of MRI and the invasion of myometrium); (b) flow voids sign present vs. absent in parametrium and myometrium, respectively.
On the other hand, reader 1 and reader 2 recorded the short diameter of the most prominent flow void (hereinafter referred to as fv-D) independently. The method of fv-D measurement is as follows: we chose the plane in which flow voids were the most prominent in the parametrium, then fv-D was measured in the coronal, sagittal and axial planes and then averaged and recorded. The same method was applied for measuring fv-D in the myometrium. If there was no flow void sign in the myometrium or/and parametrium, then we recorded the fv-D value as 0 mm. The measurements obtained for fv-D are shown in Figure 2.
To assess interobserver reliability, the fv-D measurements were performed in a blinded fashion by reader 1 and reader 2. To evaluate intraobserver reliability, the reader 1 completed the first analysis of all images and then repeated the measurements a week later.
On DSA, as a reference standard, the presence of uterine AVMs was confirmed if there were one or more draining veins on early arterial-phase images of the angiography [17,32]. The blood supply arteries and drainage veins as well as the relationship between pelvic blood vessels were evaluated.
Statistical analyses were performed with IBM SPSS Statistics version 25.0 and medcalc 19.11. Descriptive statistics included the mean and standard deviation (SD) for normally distributed variables. Categorial variables are expressed as numbers and probabilities. Fisher’s exact probability was used to compare the probability of flow void signs between patients with uterine AVMs and those without uterine AVMs. Additionally, the independent sample t-test was used to compare fv-D values in different locations between patients with and without uterine AVMs. Receiver operating characteristic (ROC) curves were plotted for fv-D in the myometrium and parametrium. The area under the curve (AUC), sensitivity, specificity, positive predictive value, negative predictive value, Youden index and cut-off value were calculated to determine which diagnostic indicator was superior. The cut-off value was determined based on the Youden index. Bland-Altman analysis was used to further assess intra- and interobserver agreement by calculating the bias (mean difference) and the 95% limits of agreement (1.96 standard deviations from the difference). Intraclass correlation coefficients (ICCs) were calculated to assess intraobserver and interobserver reliability using single measures for absolute agreement in a two-way mixed model. A two-tailed P value <0.05 was considered statistically significant.