The accuracy of ultrasound compared to magnetic resonance imaging in the diagnosis of deep infiltrating endometriosis: A narrative review

Endometriosis is defined as the presence of endometrium‐like epithelial cells and/or stroma located outside the uterus, generally with associated inflammatory response. Commonly located on the ovaries and peritoneum, endometriosis can also infiltrate the retroperitoneal space and pelvic organs, known as deep endometriosis (DE). This narrative review aims to compare the accuracy of transvaginal ultrasound (TVUS) and magnetic resonance imaging (MRI) for detecting deep endometriosis. A literature search was performed mid‐2021 and again in December 2022 using PubMed, SAGE and Wiley databases, and limited to studies published between 2011 and 2021. Key words deep infiltrating endometriosis, transvaginal ultrasound, magnetic resonance imaging and diagnostic accuracy identified 16 studies which addressed the aim. The authenticity and reliability were determined by the Standards for Reporting of Diagnostic Accuracy. When the studies were analysed, the diagnosis of DE mean sensitivity values ranged from 40% to 100% for TVUS and 50% to 100% for MRI. Mean specificity values ranged from 81.5% to 99.9% and 84.4% to 97.5%, respectively. This indicates that both MRI and TVUS have similarly accurate performance in the detection of DE. Endometriosis detection is possible with TVUS and MRI, however accuracy is dependent on the location and severity as well as limitations due to operator and image interpretation reliability. In the presence of a correctly performed dynamic TVUS examination for the diagnosis of DE, MRI adds little diagnostic value. Further research is recommended using a standardised protocol in a wider study of affected patients, including accurate clinical assessments for each location of endometriosis to inform the choice of imaging for that region.

patients. 2,3 Endometriosis is characterised as the presence of endometrium-like epithelium and/or stroma outside the uterine cavity ranging in appearance and location with associated inflammatory response causing adhesions. 4 Lesions may appear as peritoneal implants, ovarian endometriomas or encompass nodules involving organs in the pelvis. 5,6 The latter was previously known as deep infiltrating endometriosis but more commonly referred to as deep endometriosis (DE) and accounts for 15%-30% of all endometriosis scan requests. 5,6 The most frequently reported sites of DE are the vagina, bladder, uterosacral ligaments (USL), and rectosigmoid colon. 5,6 The International Deep Endometriosis Analysis (IDEA) group recently published a consensus on the terms, definitions and measurements that are recommended for use when describing the features of different phenotypes of endometriosis. 7 A systematic approach specifically for sonography is suggested when evaluating endometriosis, with consideration to the anterior compartment containing the ureters, ureto-vesical region and bladder, and the posterior compartment containing the posterior vaginal fornix, USL, rectovaginal septum and rectosigmoid colon. 7 These standardised terms have been referred to throughout this narrative review with the included relevant literature being assigned to each compartment.
Diagnostic imaging is essential for the confirmation of endometriosis as clinical and physical examination alone are of little value. 2 Multiple minimally invasive techniques have been developed for the detection of endometriosis including transabdominal ultrasound, transvaginal ultrasound (TVUS), transrectal ultrasound, magnetic resonance imaging (MRI) and less commonly, computed tomography (CT) and positron emission tomography (PET). 2,3 Undergoing laparoscopic treatment for endometriosis will yield the greatest benefit if maximum resection at first surgical intervention is achieved. 8 An accurate and detailed preoperative workup for suspected endometriosis patients, is essential for preparation and surgical mapping. 9 Thorough diagnostic imaging is crucial for predicting the potential risks and difficulties of surgical resection, the requirement for a highly experienced laparoscopic surgeon and possible involvement of a general or colorectal surgical team. 5,9 The effectiveness of surgical treatment is therefore reliant on accurate DE detection by appropriately trained medical imaging specialists and doctors. It is recommended by the World Endometriosis Society (WES) that establishing centres of experts for managing higher stage endometriosis with a reproducible and cost-effective system of triage will allow immediate comprehension of the severity of disease and appropriate counselling. 8 Determining the most effective method of pre-surgical evaluation is therefore critical for triaging patients depending on the severity of disease with appropriate preoperative mapping and counselling. 10 Limited literature is available on the diagnostic performance of CT and PET imaging for endometriosis with these modalities being compounded by high doses of ionising radiation and the increased cost of operating. 3 Although TVUS is operator dependent, it remains the preferred imaging modality due to the absence of ionising radiation, its costeffectiveness and the ease of accessibility, with MRI reserved for secondary or complementary imaging when ultrasound findings are equivocal. 11 The aim of this narrative review is to evaluate the reported diag-  Figure 1). Further analysis of the selected articles using the critical appraisal tool of 'strengthening the reporting of observational studies on epidemiology' (STROBE) eliminated three more studies from the review. 12 The Standards for Reporting of Diagnostic Accuracy was applied to determine the authenticity and reliability of the remaining studies which resulted in the inclusion of 14 appropriate articles for review. A second search was performed closer to publication (December 2022) utilising the same search terms, which identified two original research, peerreview articles that directly addressed the aim, resulting in 16 included articles ( Figure 1). 12 Data from these studies were used as the primary sources of information for the descriptive narrative in this review.

| RESULTS
The articles included have publication dates ranging from 2013 to 2022, with the only included article published in 2012 being one of only three prospective studies to directly compare TVUS with MRI, which directly addressed the aim of this narrative review. 11 Of the 16 studies reviewed (Table 1), 13 provided sensitivity and specificity results across a range of locations within the pelvic cavity (  sensitivity than TVUS in diagnostic performance across all pelvic compartments while TVUS had a higher specificity than MRI.

| Diagnostic performance: sensitivity and specificity
From the literature reviewed, TVUS had a higher sensitivity than MRI for diagnosing DE involving the vagina, ovaries and USL regions, and a higher specificity than MRI in excluding DE involving the bladder, vagina, rectovaginal and ovarian regions (Table 2). 5,11,16,[18][19][20] Both MRI and TVUS demonstrated similar diagnostic performances across the pouch of Douglas (POD) and parametrium regions. 13,16,18,20,23 When assessed in each compartment, the mean ranges for TVUS sensitivity were lower overall in comparison to MRI. Despite TVUS not being as effective as MRI at detecting DE, it was superior at excluding the disease with a greater range of mean specificities across all compartments.

| Diagnostic performance: positive and negative predictive values
The positive and negative predictive values of TVUS and MRI across the data appeared similar. As seen in Table 2, the percentage of patients with positive pathological findings in the bladder and vagina was high and relatively equal for both modalities. 5,11 Similarly, the percentage of patients with a negative predictive value (NPV) for DE located in the bladder and vagina were identical. 5,11 Transvaginal ultrasound demonstrated a higher NPV and positive predictive value (PPV) than MRI in the ovaries and rectovaginal regions, while MRI demonstrated a higher PPV and NPV in the USL, rectosigmoid and rectocervical regions. 5,11,14,17,18,21 Despite TVUS having a relatively high PPV in the POD and parametrium regions, a comparison was unable to be performed due to the limited MRI data available on PPV and NPV in this region.

| Reference standards and protocols
The majority of the reviewed studies used both intraoperative and histological results as the reference standard as seen in Figure 2, which demonstrates the diagnostic accuracy of TVUS and MRI detection of DE when compared to the surgical results provided. 6,19 The accuracy of TVUS and MRI based on surgical outcomes were comparable across both modalities with results collectively reported, with a mean accuracy of 72.8% for ultrasound and 73.9% for MRI. Transvaginal ultrasound is advantageous due to its ability to use real-time imaging to assess pelvic organ mobility, known as dynamic imaging. The literature suggested that when mobility of the ovaries on dynamic TVUS imaging cannot be demonstrated, this may indicate adhesions and the potential presence of DE cannot be excluded. 6,19 Ovarian the evaluation protocol based on the IDEA consensus. 19 The reported accuracy of these studies were similar to other research that failed to apply a standardised protocol, which would have allowed for a more coherent analysis and further, correlation between the results of each modality and surgical findings. As such, creating an MRI protocol based on the current IDEA consensus, as proposed by Indrielle-Kelly et al. 19 may not affect the accuracy of imaging but comparison between reporting may be more easily achieved.
Another study assessed the inter-observer reliability by reviewing the imaging of endometriosis assessment under MRI. 16 Images were reviewed by two radiologists with 10-15 years expertise in body MRI interpretation. Radiologists were blinded to the laparoscopic and histopathological results with any discrepancies being resolved by consensus. 16 Three nodules required re-evaluation by two radiologists to resolve discrepancies caused by prominent susceptibility artefacts on MRI. 16 These artefacts caused by intestinal gas in the pelvic region increased with 3T MRI when compared to 1.5T. 16

| Operator dependency
When performed by an experienced sonographer, TVUS was accurate, reproducible and minimally invasive in the detection of DE. 17 According to a study by Savelli et al., 9 when TVUS was performed by experienced operators, the sensitivity, specificity, PPVs and NPVs were significantly higher than when TVUS was performed by less experienced operators. The accuracy of the preoperative workup for suspected endometriosis patients was crucial for predicting the risks and difficulties, the requirement for a highly experienced laparoscopic surgeon and possible involvement of a general or colorectal surgical team. 9  This meta-analysis highlighted the advantage of highly experienced operators and demonstrated that discrepancies were resolved with consensus, however the dependency on the performing sonographers', reporting physicians' and surgeons' skill remained a limitation. 27 A larger prospective study utilising inter-observer reliability is required to validate these assumptions, though it should be noted that not all examiners will have a high experience in transvaginal ultrasound, and as such any such study may not be representative of the scanned population.

| Deep endometriosis
As previously identified, this narrative review agrees that DE is most commonly identified on the USLs, rectosigmoid colon, vagina, and bladder. 5,11 These results correlate with similar findings in a recently published cross-sectional analysis studying the anatomical distribution of endometriosis in symptomatic patients. 31 Surgical laparoscopic resection remains the primary method of treatment, however due to the associated risks of surgery, preoperative assessment is deemed important for appropriate planning. 5 In a study by Alborzi et 10,13,14,16,17,19,22,27 While multiple studies included large patient pools consisting of high and low risk populations, such as those presenting with clinical symptoms or chronic pelvic pain, future research is needed to assess the diagnostic accuracy of TVUS in an undiagnosed population. 5,6,11,15,18,20,27 With the exception of one study, our narrative review included original research published within the last 10 years whereas even the more recent systematic reviews and meta-analysis included articles published prior to 2013. 27

CONFLICT OF INTEREST STATEMENT
None.