Ultrasound and MRI: what they can and cannot show

Ultrasound and MRI can contribute important information in the assessment of endometriosis and adenomyosis, but neither has one accuracy level that applies to every target and setting. A scan may assess an ovarian cyst, deep disease at a particular site, superficial disease, or features in the uterine muscle. The protocol, reporting language, operator, study population, and reference standard all affect the evidence. A useful explanation therefore asks what was sought and what was observed before treating a report as a complete answer to pelvic pain. ^{[1][2][3][13][14]}

A test needs a named target

The IDEA consensus was developed to standardize ultrasound terms, definitions, and measurements for endometriosis. Its authors described difficulty comparing studies that used different terms for the same structures and locations. The aim was consistent description of location and extent, supporting communication and future research. This is an essential first step in understanding accuracy: the target being assessed must be clear before results can be meaningfully compared. ^[1]

“Endometriosis detected” can conceal several different questions. The Cochrane diagnostic review separately considered overall pelvic endometriosis, ovarian endometriomas, deep endometriosis, and disease at particular anatomical sites. It also distinguished tests intended to replace surgery from tests intended to help decisions about surgery. These are different purposes, so a test can perform usefully for one without satisfying criteria for another. ^[3]

A single overall percentage can erase those distinctions. It may combine people with different target findings or summarize a study with a particular case mix. The first interpretive step is therefore not deciding whether the number looks high. It is identifying the target, the population, and the comparison that produced it. The diagnostic reviews explicitly separate categories because those differences change what the evidence means. ^[3][4][5]

Routine ultrasound and focused observations

The Society of Radiologists in Ultrasound consensus explains that endometriosis may not be appreciated during routine transvaginal ultrasound beyond an ovarian endometrioma. Its multidisciplinary panel proposed additions to assessment, including observations of the posterior compartment, relative organ positions, and the uterine sliding sign. The document also proposes a reporting approach. These are concrete changes in what is observed, rather than a claim that the word “ultrasound” already specifies a uniform examination. ^[2]

The statement was developed through literature review and a modified Delphi process. A consensus can recommend a technique and terminology, but it does not measure how widely the technique has been adopted or what happens in every service. The recommendations should therefore remain distinct from claims about real-world implementation or guaranteed earlier diagnosis. The existence of a protocol does not establish that a particular report used it. ^[2]

This explains why the details of an examination can matter more than its broad name. A report may describe ovarian structures, movement, deep sites, or only a subset of those observations. The IDEA and SRU documents support more consistent description, while leaving the actual scope of an individual examination to its protocol and record. A public article cannot infer those details from a short phrase alone. ^[1][2]

Ovarian endometriomas are one detection problem

A systematic review of endometrioma imaging included 16 prospective studies and 1,976 participants with suspected endometriosis. It examined transvaginal ultrasound and MRI against surgical visualization, with or without histopathological confirmation. The pooled estimates indicated high accuracy for identifying endometriomas. These findings concern ovarian endometriomas specifically; they should not be repeated as the accuracy of imaging for every form of endometriosis. ^[5]

The pooled sensitivity estimates were 89% for ultrasound and 94% for MRI, with specificity estimates of 95% and 94%, respectively. However, the studies had high or unclear risk of bias, and only two directly compared the modalities in the same participants. That means a simple comparison of the pooled numbers is not a definitive head-to-head test of which modality is better for all patients. ^[5]

The reference standard is another part of the result. Some studies used surgical visualization with tissue confirmation, while others did not require both. Differences in how a finding is confirmed can affect comparison across studies. The review's favorable accuracy findings remain useful, but the uncertainty and target definition belong beside them. They cannot establish what a negative scan means for superficial disease or an unrelated cause of pelvic pain. ^[5]

Deep disease varies by location

A head-to-head meta-analysis of deep endometriosis imaging included six studies and 424 participants with clinical suspicion of deep disease. Each participant received both ultrasound and MRI, with surgery used as the reference standard. The review considered sites including the rectosigmoid, uterosacral ligaments, and rectovaginal septum. It found broadly similar performance between the modalities at those sites, with wide confidence intervals and moderate or high heterogeneity for many estimates. ^[4]

The site-specific approach is important. For rectosigmoid disease, pooled sensitivity was 85% for both modalities, while estimates differed at other locations. These numbers do not create a universal “deep endometriosis sensitivity.” A test can perform differently according to the anatomical target, and a study may include relatively few cases at one site. The uncertainty expressed by confidence intervals is therefore part of the evidence rather than an optional detail. ^[4]

A same-participant comparison reduces one source of difference between modalities, because both tests encounter the same recruited group. It does not remove every limitation. The studies still involved people already suspected of having deep disease and proceeding to surgery. Their results cannot automatically be applied to general screening or used to infer the probability of disease in an individual with a different clinical context. ^[4]

Superficial disease is harder to exclude

A retrospective London study examined a targeted ultrasound approach to the peritoneum of the pouch of Douglas. It included 100 people who had scans by one clinician trained in endometriosis assessment and later underwent laparoscopy. Of 33 people with superficial disease at surgery, 17 were identified on ultrasound. Sensitivity was 51.5%, while specificity was 94.0%. The study demonstrates possible detection while also showing why a negative result can leave uncertainty. ^[6]

The distinction between sensitivity and specificity is central here. Sensitivity describes the proportion of reference-positive cases detected; specificity describes the proportion of reference-negative cases correctly classified. High specificity does not compensate for missed cases by making every negative result conclusive. The two measures answer different questions. In this study, both the positive findings and the limits of sensitivity belong in the explanation. ^[6]

Another prospective study at McMaster included 125 people aged 18–50 with suspected endometriosis who underwent standardized ultrasound and laparoscopy. It again found stronger performance for ovarian and deep disease than for superficial disease, with superficial sensitivity ranging from 4.0% to 43.5% across assessed sites. Differences between the two studies do not mean one number must be the universal truth. They show the relevance of target definitions, technique, sites, and recruited populations. ^[7]

Prospective validation is a step, not universal implementation

A prospective study at a Qingdao hospital evaluated SRU-based ultrasound assessment in 288 patients with clinical suspicion of deep endometriosis. It compared imaging with surgery and reported sensitivity of 87.4% and specificity of 76.8% for deep disease. The investigators also examined associations between indirect signs and the locations of lesions. This adds performance evidence beyond the original consensus recommendation. ^[8]

The study population and setting still matter. Everyone entered through clinical suspicion, and surgical findings supplied the comparison. These are not general-population screening results. Associations between indirect signs and lesion sites are group-level findings from that study; they cannot be used by a website to assign a personal probability or determine whether a particular report establishes disease. The validation is valuable within its defined question. ^[8]

It also measures diagnostic performance, rather than the whole effect of introducing the approach into routine services. Shorter diagnostic intervals, access, patient understanding, or long-term outcomes would require their own measurements. A protocol can have promising accuracy while those broader questions remain unanswered. Keeping the outcome named prevents a technical advance from being promoted into a claim about every part of care. ^[2][8]

What the sliding sign does and does not mean

A systematic review examined the sliding sign in relation to two specific targets: obliteration of the pouch of Douglas and bowel involvement. It included eight studies, with different numbers of participants available for the two analyses, and used laparoscopy as the reference standard. The targets were anatomical findings in people with suspected endometriosis, not the presence of every possible form of the condition. ^[9]

The pooled sensitivity and specificity for pouch-of-Douglas obliteration were 88% and 94%. For bowel involvement, they were 81% and 95%, with high heterogeneity in the bowel analysis. These estimates support usefulness for those questions while retaining uncertainty and variation. A sliding-sign result should not be converted into a universal endometriosis diagnosis or exclusion simply because it is one part of a focused assessment. ^[9]

The review also illustrates why one maneuver should not stand for the entire examination. Its results describe how that observation related to named targets in included studies. Other anatomical sites, superficial lesions, or uterine-muscle features pose different questions. A clear report can include the observation and its interpretation without implying that every unmentioned target has been evaluated. ^[9]

Mapping with a common classification

A prospective multicenter study evaluated ultrasound mapping using the #Enzian classification in 745 women undergoing radical surgery for deep endometriosis at tertiary centers. The classification organizes findings by compartments and other locations. Researchers compared preoperative ultrasound with surgical assessment, including lesion locations, grades, and adhesions. This is evidence about structured mapping in a selected specialist population. ^[10]

Performance differed between compartments. The study found that identifying and grading tubo-ovarian adhesions was more difficult than assessing several lesion categories. The finding matters because agreement about “endometriosis” overall can hide uneven performance in the components needed for planning. A classification makes those components explicit; it does not make every one equally easy to assess. ^[10]

The size and multicenter design broaden the evidence, but the participants were still selected for specialist surgery. Results should not be presented as the accuracy of every routine scan. The study supports a common language for comparing ultrasound and operative mapping within its setting, while leaving transfer to other populations and services as a separate question. ^[10]

MRI indications and the question being asked

An ESUR consensus addresses MRI indications, reporting, and classification. Its recommendations include situations with inconclusive or negative transvaginal ultrasound despite symptoms, assessment before an intervention, and persistent symptoms after surgery. These are expert recommendations concerning the role of MRI within a pathway. They are not a statement that every person with pelvic pain should receive the examination or that the result will resolve every uncertainty. ^[11]

The panel used a Delphi process involving 20 experts from ten countries. It emphasized standardized reports organized by anatomical compartments, with optional drawings to aid communication. That focus is practical: a report can provide more useful mapping information than a short positive or negative label. The consensus is evidence-informed guidance, rather than a new cohort establishing one universal MRI accuracy estimate. ^[11]

A companion ESUR document addresses protocol and vocabulary. It describes sequences and systematic anatomical assessment, again emphasizing compartment-based analysis. The two papers arise from a related expert process and should not be counted as independent replications of a diagnostic trial. They explain different aspects of a standardized approach, not two separate demonstrations that every implementation will achieve the same result. ^[11][12]

MRI protocol and report content are part of the evidence

The protocol consensus describes how different MRI sequences and anatomical observations contribute to mapping. It discusses describing endometriomas by characteristics such as number, location, and laterality, and describing deep findings by site and dimensions. These details show why a report contains more information than the name of the modality. They also show why a generic label cannot establish that every relevant compartment was assessed in the same way. ^[12]

The document includes preparation and technical recommendations for professionals. This article does not turn them into instructions for patients to prepare independently or use medication. Their relevance here is methodological: studies and protocols may not use identical acquisition conditions. Comparing accuracy or reports without recognizing those conditions can make different examinations appear more interchangeable than the evidence supports. ^[12]

Mapping also needs to remain separate from predicting symptoms. A detailed description can help communication and planning, but it does not directly measure pain or daily function. The consensus emphasizes communication with surgeons and patients. That purpose is compatible with leaving other clinical questions unresolved, rather than expecting an anatomical map to provide a complete account of a person's experience. ^[11][12]

Adenomyosis asks about the uterine muscle

The revised MUSA consensus separates direct and indirect ultrasound features of adenomyosis. Direct features include findings such as myometrial cysts and bright tissue islands; indirect features include changes in uterine shape, wall thickness, shadowing, vessels, and the junctional zone. These are observations in the uterine muscle, distinct from many of the targets in endometriosis-focused mapping. A pelvic scan may address both questions, but the frameworks are not interchangeable. ^[13]

A head-to-head review compared ultrasound and MRI in the same participants across six adenomyosis studies, totaling 595 women. Hysterectomy pathology supplied the reference standard. Pooled sensitivity and specificity were 75% and 81% for ultrasound, and 69% and 80% for MRI. The review found no statistically significant difference, but high heterogeneity and patient-selection concerns limit broad conclusions. ^[14]

The finding supports similar performance within those studies, rather than a universal rule that MRI is superior, ultrasound is sufficient, or either test settles every case. A selected hysterectomy population is not equivalent to everyone with bleeding or pain. The target, reference standard, and recruitment route remain part of the result, just as they do in endometriosis imaging research. ^[14]

Directed assessment and ordinary reports can differ

A study of clinical ultrasound and MRI reports compared recorded adenomyosis findings with later hysterectomy pathology. It found lower detection in routine practice than expected from focused research assessments. When radiologists directly reevaluated MRI for adenomyosis, accuracy improved in the study dataset. This makes the distinction between an available image and an assessment specifically directed at a question concrete. ^[15]

The study does not establish that every unmentioned condition was missed or that every routine report should be discounted. It came from a selected institutional surgical population. Its contribution is evidence that communication about the assessment question and reporting practices can affect the record. A careful public summary retains this scope rather than assigning the study's percentages to all imaging everywhere. ^[15]

Consensus, accuracy, and clinical outcomes are different evidence

An international multisociety consensus produced 20 statements about imaging for deep endometriosis and classification. Agreement varied: most statements reached strong or moderate agreement, while one remained in equipoise. The process illustrates both the value of coordinated guidance and the limits of consensus. The number of participating societies is not a count of independent diagnostic experiments. ^[16]

The Cochrane review provides another kind of evidence: an assessment of diagnostic studies, including their methodological quality and varied targets. Its literature search ended years before several later studies discussed here. It should be read in that historical context, rather than as a permanent claim that noninvasive diagnosis is impossible. Later consensus and performance studies address developments without erasing the earlier review's concerns about study design and heterogeneity. ^{[3][7][8][16]}

Why confidence intervals belong beside accuracy estimates

The head-to-head review of deep endometriosis reported wide confidence intervals for several estimates. For example, its pooled MRI sensitivity for rectovaginal-septum disease was 66%, with a 95% confidence interval from 51% to 79%; the ultrasound estimate was 59%, with an interval from 26% to 86%. Reporting only 66% versus 59% would make the comparison appear more settled than the intervals and study heterogeneity support. ^[4]

A confidence interval expresses statistical uncertainty in an estimate under the analysis used. It is not a range of possible symptoms for a person, and it is not the probability that a reader has disease. The review calculated these intervals for study-level diagnostic measures. Their width is relevant because the evidence does not locate the underlying performance with equal precision at every anatomical site. ^[4]

The endometrioma review provides a second example. Its pooled MRI sensitivity estimate was 94%, but its confidence interval extended from 74% to 99%. The review also found only two studies with direct same-participant comparisons between MRI and ultrasound. Those qualifications prevent a small difference in headline percentages from becoming proof that one modality is always better. Both the number and the uncertainty are part of the finding. ^[5]

The practical implication for reading research is to retain the target and the interval together. “High accuracy” is a summary phrase; it should not replace the actual evidence when comparing studies or explaining limits. These reviews describe performance across recruited groups and reference standards. They do not authorize a personal conclusion from a percentage detached from its denominator, protocol, or anatomical question. ^[4][5]

What this does not tell you

This article cannot interpret a personal scan, decide whether a test is appropriate, establish a diagnosis, or assign an individual probability. The sources include consensus documents, specialist cohorts, and systematic reviews with differing methods and limitations. A high accuracy estimate for one target cannot be generalized to every site, protocol, operator, or population. A comparison between modalities also does not determine the next step for an individual. ^{[3][4][5][8][14]}

A negative report needs its scope preserved. It does not automatically exclude every form of endometriosis, adenomyosis, or every other cause of pain; equally, a detection limit does not prove a hidden diagnosis. The appropriate account separates the question asked, observations made, interpretation offered, and uncertainty that remains. That keeps imaging useful without asking it to promise more than its evidence can support. ^{[6][7][14][15]}