{"paper_id":"8531f69c-c523-41d8-817e-8bcf6bf58d6a","body_text":"Review began\n 10/17/2026 \nReview ended\n 10/28/2026 \nPublished\n 10/31/2026\n© Copyright \n2026\nAhmed et al. This is an open access article\ndistributed under the terms of the Creative\nCommons Attribution License CC-BY 4.0.,\nwhich permits unrestricted use, distribution,\nand reproduction in any medium, provided\nthe original author and source are credited.\nDOI:\n 10.7759/cureus.118898\nDiagnostic Accuracy of Transvaginal Ultrasound\nVersus Magnetic Resonance Imaging for Deep\nInfiltrating Endometriosis: A Histopathologically\nCorrelated Prospective Study\nDoaa Ahmed \n, \nAmina \nIqbal \n, \nGauri \nSreeram \n, \nUrooj Anwar \n, \nNida \nWaqas \n, \nTitilayo \nOlawunmi \n,\nGyanendra KC \n, \nVanessa Colaco \n, \nRachel Marsha Manoj Stanley \n, \nNazia Sumbul Mahir \n, \nSaadia Hassan \n1.\n Obstetrics and Gynaecology, Guy's and St Thomas' NHS Foundation Trust, London, GBR \n2.\n Radiology, Cheema Heart\nComplex Gujranwala, Gujranwala, PAK \n3.\n Radiology, Mallige Hospital, Bengaluru, IND \n4.\n Obstetrics and Gynaecology,\nAkhtar Saeed Medical and Dental College, Lahore, PAK \n5.\n Obstetrics and Gynaecology, Khokhar Hospital, Mandi\nBahauddin, PAK \n6.\n Medicine, Ekiti State University, Ekiti, NGA \n7.\n Radiology, KIST Medical College and Teaching\nHospital, Imadol, NPL \n8.\n Obstetrics and Gynaecology, Oxford University Hospitals NHS Foundation Trust, Oxford, GBR \n9.\n Emergency, Glenfield Hospital, Leicester, GBR \n10.\n Obstetrics and Gynaecology, Luhansk State Medical University,\nLuhansk, UKR \n11.\n Gynaecology, Rawalpindi Medical University, Rawalpind, PAK\nCorresponding author: \nSaadia Hassan, \nhassansaadia695@gmail.com\nAbstract\nBackground\nDeep infiltrating endometriosis (DIE) requires accurate anatomical assessment. This study prospectively\ncompared transvaginal ultrasound (TVUS) and magnetic resonance imaging (MRI) with histopathology as\nthe reference standard.\nMethods\nIn this prospective paired diagnostic-accuracy study, 350 women with clinically suspected DIE underwent\nstandardized TVUS and pelvic MRI, followed by laparoscopy and histopathological assessment. We\ncomputed overall and anatomical site-specific sensitivity, specificity, positive predictive value (PPV),\nnegative predictive value (NPV), accuracy, and likelihood ratios. We compared paired classifications using\nMcNemar’s test.\nResults\nHistopathology confirmed DIE in 213 (60.9%) participants. TVUS demonstrated 89.2% sensitivity, 86.9%\nspecificity, 91.3% PPV, 83.8% NPV, and 88.3% accuracy. Corresponding MRI estimates were 91.5%, 92.7%,\n95.1%, 87.6%, and 92.0%, respectively. TVUS showed higher sensitivity for ureteric disease, whereas MRI\nshowed higher specificity and accuracy for several other anatomical sites. Paired classifications showed no\nsignificant differences (McNemar’s \nχ\n² = 0.066, P = 0.798), although the counts differed. Interobserver\nagreement was substantial to near-perfect for most sites.\nConclusion\nBoth standardized TVUS and MRI demonstrated high diagnostic accuracy for DIE. MRI performed better\nnumerically for several measurements, but the paired analysis did not show statistically significant\ndifferences in overall classification. Expert TVUS may serve as an effective first-line examination, with MRI\nproviding complementary information for complex, extensive, or equivocal disease and preoperative\nmapping.\nCategories:\n Obstetrics/Gynecology, Radiology\nKeywords:\n deep infiltrating endometriosis, diagnostic accuracy, histopathology, magnetic resonance imaging,\ntransvaginal ultrasound\nIntroduction\nEndometriosis is a chronic inflammatory gynecologic disorder that affects 2-10% of the general female\npopulation, reduces quality of life, and is associated with dysmenorrhea, dyspareunia, pelvic pain, and\ninfertility \n[1]\n. Deep infiltrating endometriosis (DIE) is defined by lesions that penetrate deeply into pelvic\ntissues, including the uterosacral ligaments, rectovaginal area, bowel, bladder, and other pelvic organs. The\nlocation and extent of disease matter for preoperative evaluation, surgical complexity, and treatment\nplanning \n[2,3]\n. Accurate preoperative identification and anatomical classification of DIE are therefore\nimportant for patient assessment and appropriate planning of surgical treatment strategies \n[4]\n.\n1\n2\n3\n4\n5\n6\n7\n8\n9\n10\n11\n \nOpen Access Original Article\nHow to cite this article\nAhmed D, Iqbal A, Sreeram G, et al. (October 31, 2026) Diagnostic Accuracy of Transvaginal Ultrasound Versus Magnetic Resonance Imaging for\nDeep Infiltrating Endometriosis: A Histopathologically Correlated Prospective Study. Cureus 18(10): e118898. \nDOI 10.7759/cureus.118898\n\nNoninvasive imaging may be used as part of the initial diagnostic approach for patients with suspected\nendometriosis, as surgical evaluation with histopathological examination is not necessary to make the\ndiagnosis \n[5]\n. Noninvasive imaging has thus become the key factor in the assessment of patients with\nsuspected DIE \n[4]\n. Transvaginal ultrasound (TVUS) is widely available, relatively inexpensive, and can be\nperformed using a systematic, endometriosis-specific protocol that provides detailed assessment of pelvic\nstructures \n[6]\n. Magnetic resonance imaging (MRI), however, allows excellent soft-tissue characterization and\ncomprehensive multiplanar evaluation of the pelvis, with special benefits for extensive, multicompartment,\nor anatomically complex disease \n[7,8]\n. Nevertheless, reported diagnostic efficacy varies widely across\nmodalities, depending on lesion localization, imaging protocols, operator experience, and patient selection\n[9]\n.\nDespite growing evidence supporting both TVUS and MRI, an important clinical question remains\nunresolved: does MRI provide superior diagnostic accuracy to appropriately performed TVUS for detecting\nDIE, and are any differences in diagnostic performance modality- or site-specific? \n[10]\n. Previous\ninvestigations have used diverse imaging modalities and methods to define and assess anatomical DIE, with\nvaried study designs and reference standards for diagnostic confirmation ranging from surgical to\nhistological \n[8,9]\n. Moreover, overall diagnostic assessment may mask clinically relevant differences between\npelvic compartments \n[11]\n. Within each patient, direct prospective comparison using a uniform imaging\nstrategy and histopathology as the reference standard can provide a more meaningful assessment of each\nmodality’s relative diagnostic value \n[11]\n.\nTherefore, this prospective paired diagnostic-accuracy study aimed to compare the diagnostic performance\nof TVUS and MRI for detecting DIE, using histopathological examination as the reference standard. The\nprimary objective was to evaluate and compare the sensitivity, specificity, positive predictive value (PPV),\nnegative predictive value (NPV), and overall diagnostic accuracy of both modalities for detecting\nhistopathologically confirmed DIE. Secondary objectives were to compare their diagnostic performance\nacross predefined anatomical sites and assess interobserver agreement in site-specific disease detection.\nRather than presuming the superiority of either imaging modality, this study sought to characterize their\ncomparative diagnostic performance and identify potential differences in their strengths across anatomical\nsites. The findings may help inform the selection and complementary use of these modalities to optimize\npreoperative imaging assessment and surgical planning in women with suspected DIE.\nMaterials And Methods\nStudy design and setting\nWe conducted a prospective paired diagnostic-accuracy study to compare the diagnostic performance of\nTVUS and MRI for detecting and anatomically mapping DIE. The paired design allowed each participant to\nundergo both index imaging examinations (TVUS and MRI) before the same surgical and histopathological\nreference-standard assessment, enabling direct within-participant comparison of the two imaging\nmodalities. The study was conducted in the gynecology unit of Holy Family Hospital, Rawalpindi, Pakistan,\nwith dedicated gynecological ultrasound units, pelvic MRI, laparoscopic surgery for gynecological\nconditions, and histopathology facilities. We recruited participants consecutively between January 2026 and\nAugust 2026.\nStudy population\nThe study population comprised adult women with clinically suspected endometriosis or DIE who were\nscheduled for laparoscopic surgical evaluation. We screened women presenting to the\ngynecology/endometriosis service during the recruitment period for eligibility. To limit selection and\nspectrum bias, we recruited women with suspected disease rather than only those with a diagnosis,\nproviding a clinically enriched population of women referred for specialist evaluation of suspected\nendometriosis.\nDuring the study period, we assessed 412 women for eligibility. Of these, 62 were ineligible: 34 did not meet\nthe inclusion criteria, 16 refused to participate, and 12 had incomplete data. The remaining 350 women were\nselected for the study and underwent both TVUS and MRI, followed by laparoscopy and histopathological\nexamination. All 350 participants completed the diagnostic pathway and constituted the final analytic\ncohort. Figure \n1\n presents the participant flow.\n \n2026 Ahmed et al. Cureus 18(10): e118898. DOI 10.7759/cureus.118898\n2\n of \n16\n\nFIGURE\n 1: Participant flow diagram for the prospective paired\ndiagnostic-accuracy study\nInclusion and exclusion criteria\nWomen were eligible if they were aged 18 years or older, had clinical suspicion of endometriosis or DIE, were\nscheduled for definitive surgical evaluation, and underwent both preoperative TVUS and MRI before surgery.\nClinical suspicion was based on compatible symptoms, including dysmenorrhea, chronic pelvic pain, deep\ndyspareunia, dyschezia, dysuria or cyclical urinary symptoms, infertility, or cyclical bowel symptoms. All\nparticipants underwent surgical evaluation and provided written informed consent. All participants\nunderwent laparoscopic surgical assessment with tissue sampling for histopathological examination.\nWomen were excluded if they were pregnant, had known pelvic malignancy, had undergone previous\nhysterectomy, had a contraindication to MRI, had an incomplete MRI examination or inadequate TVUS\nexamination, underwent surgery before completion of both imaging examinations, had missing essential\nimaging information, were unable to undergo the planned surgical evaluation, or lacked relevant\nhistopathological assessment required for the primary lesion-based analysis. Women who refused to\nparticipate were also excluded. Previous endometriosis surgery was not an automatic exclusion criterion;\ninstead, it was recorded as a clinical characteristic for subsequent secondary or sensitivity analyses.\nRecruitment and study procedures\nWe recruited eligible women sequentially and conducted a standardized clinical assessment before imaging.\nWe documented demographic characteristics, anthropometric measurements, gynecological history,\nsymptom profile, previous endometriosis diagnosis or surgery, previous pelvic surgery, hormonal treatment,\nand relevant gynecological conditions using a standardized data collection form.\nAll participants underwent both TVUS and MRI before surgical intervention. We performed the two imaging\nexaminations within a predefined interval, preferably within four weeks of each another, and completed\nboth before surgery. We documented any deviation from the predefined interval and considered it in\nsensitivity analyses, where appropriate.\nIndex tests\nThe two index tests were standardized TVUS and pelvic MRI. We compared the modalities in terms of their\nability to detect histopathologically confirmed DIE. Both examinations involved systematic anatomical\nassessment rather than evaluation restricted to a single suspected lesion.\nTVUS\nWe performed TVUS using a GE Voluson E10 ultrasound system equipped with an RIC5-9-D 5-9 MHz\nendovaginal probe (Wipro GE Medical Device Manufacturing Private Limited, Bengaluru, India). An\nexperienced gynecological sonologist with at least three years of experience in gynecological ultrasound and\nspecialized experience in endometriosis imaging performed the examination. The examiner was trained in\n \n2026 Ahmed et al. Cureus 18(10): e118898. DOI 10.7759/cureus.118898\n3\n of \n16\n\nthe International Deep Endometriosis Analysis (IDEA) approach. The TVUS examiner independently\ninterpreted and recorded the ultrasound findings and was blinded to the MRI findings during image\ninterpretation. The examiner systematically assessed pelvic anatomy using a standardized, IDEA-based\nexamination.\nThe examination first assessed uterine position, uterine dimensions, adenomyosis, myometrial\nabnormalities, and endometrial appearance. The ovaries were then examined for abnormalities and\nendometriomas, assessing laterality, number, and largest dimensions. We evaluated pelvic organ mobility\nusing the sliding sign and assessed uterine mobility relative to the bladder and rectum, categorizing it as\npositive or negative. We categorized the pouch of Douglas as not obliterated, partially obliterated, or\ncompletely obliterated.\nWe systematically examined the posterior compartment, including the uterosacral ligaments, torus uterinus,\nrectovaginal septum, vagina, rectum, rectosigmoid, and pouch of Douglas. We examined the anterior\ncompartment for abnormalities involving the urinary bladder, vesicouterine space, and distal ureters.\nFor each suspected DIE lesion, we recorded its presence or absence, anatomical site, laterality, number, and\ndimensions. We documented lesion dimensions as length × width × depth in millimeters. Additionally, where\npossible, we recorded the number of nodules, lesion length, and distance from the anal verge in\nrectosigmoid lesions. We also documented associated organ involvement.\nMRI\nPelvic MRI was performed on a 3-T Siemens MAGNETOM Skyra MRI system (Siemens Healthineers,\nForchheim, Germany) with a pelvic phased-array coil. All patients underwent the same preimaging\npreparation: fasting, bowel preparation, moderate bladder filling, and administration of an antiperistaltic\nagent before imaging. MRI acquisitions comprised high-resolution, multiplanar T2-weighted sequences in\nthe sagittal, axial, and oblique planes, along with T1-weighted and fat-suppressed T1-weighted sequences.\nDiffusion-weighted images were obtained using b-values of 50, 400, and 800 s/mm². Images were acquired\nwith a slice thickness of approximately 3-4 mm and a field of view of approximately 20-24 cm, with small\ninterslice gaps where applicable. Gadolinium contrast was not routinely used to assess DIE. Additional renal\nimaging was performed to evaluate possible urinary tract involvement.\nAn experienced pelvic radiologist with more than three years of experience in pelvic MRI, including\nexpertise in gynecological imaging and endometriosis, interpreted the MRI examinations. The radiologist\nsystematically assessed the predefined anatomical sites according to the study protocol and was blinded to\nthe TVUS findings during image interpretation. The MRI evaluation was systematic and included the\nanterior, central, posterior, and lateral pelvic compartments. The bladder, vesicouterine space, and distal\nureters were included in the anterior compartment. The central compartment comprised the cervix, vagina,\nand rectovaginal septum. The posterior compartment included the pouch of Douglas, rectosigmoid, rectum,\ntorus uterinus, and uterosacral ligaments. The lateral compartment comprised the parametrial structures,\nureters, and the pelvis sidewalls. The ovaries, endometriomas, appendix, cecum, small bowel, sigmoid colon,\nand upper abdominal and diaphragmatic regions were also evaluated when clinically indicated.\nFor each suspected lesion, the radiologist recorded its presence or absence, anatomical location, laterality,\nnumber, size, depth or extent, associated organ involvement, and evidence of obstruction, where applicable.\nWhen measurable, the lesion size and length, distance from the anal verge, and circumferential involvement\nwere recorded for rectosigmoid lesions. The MRI protocol followed European Society of Urogenital\nRadiology guidelines for MRI evaluation of pelvic endometriosis \n[12]\n.\nDefinition of DIE\nDIE was defined as endometriotic disease affecting the subperitoneal tissues, with infiltration of underlying\ntissues, typically presenting as fibrotic or nodular lesions. Imaging appearances were not used as the\ndefinitive diagnostic criterion for the primary analysis. Diagnostic classification was based on the\nprespecified histopathological reference standard.\nFor both TVUS and MRI, findings were recorded according to the presence or absence of suspected DIE, the\nanatomical site involved, and lesion dimensions, where available. The predefined anatomical sites included\nthe rectosigmoid, rectum, uterosacral ligaments, rectovaginal septum, vagina, bladder, and ureters.\nHistopathological examination served as the reference standard for diagnostic-accuracy analyses.\nHistopathology variables were coded as binary outcomes, with 1 indicating histopathological confirmation\nof DIE and 0 indicating that DIE was not recorded as histopathologically confirmed at the corresponding site.\nSite-specific imaging findings were compared with the corresponding histopathology classifications. The\nanatomical sites and imaging classification approach were defined according to the study protocol, and no\nadditional site-specific imaging classification criteria were introduced retrospectively.\nSurgical assessment\n \n2026 Ahmed et al. Cureus 18(10): e118898. DOI 10.7759/cureus.118898\n4\n of \n16\n\nExperienced gynecologic surgeons with expertise in endometriosis surgery performed all surgical\nprocedures. All patients underwent laparoscopic surgery or surgery with laparoscopic assistance, depending\non the clinical situation. The operating surgeon carefully examined the pelvis and documented the location,\nnumber, and size of suspected lesions and the organs involved. Particular attention was paid to the\nuterosacral ligaments, pouch of Douglas, rectovaginal region, rectum and rectosigmoid, urinary bladder, and\nureters. The presence and extent of adhesions and obliteration of the pouch of Douglas were also\ndocumented. The operating surgeon was not blinded to the clinical and imaging findings available before\nsurgery. We recorded operative findings independently at the time of surgical assessment.\nHistopathological reference standard\nHistopathological examination of surgically obtained tissue specimens served as the reference standard for\nthe diagnostic-accuracy analysis. All 350 patients underwent laparoscopic surgery following TVUS and MRI,\nand tissue specimens were submitted for histopathological examination. Histological assessment focused on\nthe presence of endometrial-type glands and endometrial-type stroma, with fibrosis and hemosiderin-laden\nmacrophages considered supportive findings. DIE was classified as histopathologically confirmed when\nendometrial-type glands or stroma were identified in the examined tissue, according to the prespecified\ndiagnostic criteria. We identified the anatomical site of each specimen from the corresponding surgical and\nhistopathological records and linked site-specific findings to the relevant imaging findings for diagnostic-\naccuracy analysis. Site-specific imaging findings were compared with the corresponding histopathological\nclassifications. The pathologist was blinded to the TVUS and MRI findings during histopathological\nassessment.\nBlinding and bias control\nTo reduce interpretation and review bias, the TVUS examiner was blinded to the MRI findings, and the MRI\nradiologist was blinded to TVUS findings during image interpretation. The pathologist was blinded to the\nTVUS and MRI findings during histopathological assessment. Imaging readers recorded their findings before\naccessing the corresponding results from the other imaging modality. Participants received coded\nidentification numbers, and statistical analyses were conducted using the coded dataset.\nConsecutive recruitment was used to reduce selection bias, while the inclusion of women with suspected\nrather than already confirmed DIE reduced spectrum bias. Imaging examinations were performed before\nsurgery to avoid affecting image interpretation. To minimize timing-related bias, the interval between\nimaging and surgery was defined before the study. All participants underwent the index imaging\nexaminations within four weeks before laparoscopic surgery and histopathological assessment. Participants\nunderwent the same reference-standard surgical and histopathological evaluation, where possible, to\nminimize verification bias.\nSample size and final study cohort\nThe final analytic cohort included 350 participants. In total, we assessed 412 women for eligibility and\nexcluded 62 because of ineligibility (n = 34), unwillingness to participate (n = 16), or missing data (n = 12).\nThe remaining 350 participants underwent both TVUS and MRI, followed by laparoscopic surgery and\nhistopathological examination, and were included in the final diagnostic-accuracy analysis. No formal a\npriori sample-size calculation was performed. The absence of a formal a priori sample-size calculation\nshould be considered when interpreting the precision of the diagnostic-accuracy estimates, particularly for\nanatomical sites with relatively few histopathologically confirmed cases. The final sample size was based on\nthe number of consecutive eligible participants who completed the prespecified diagnostic pathway during\nthe study recruitment period. All 350 participants underwent both index tests and the same histopathological\nreference-standard assessment; therefore, a paired design was used to compare the diagnostic performance\nof TVUS and MRI.\nStatistical analysis\nStatistical analyses were performed using IBM SPSS Statistics, version 26.0 (IBM Corp., Armonk, NY, USA).\nDescriptive statistics were used to summarize the study variables. Diagnostic performance was evaluated\nusing 2 × 2 contingency tables, with histopathology serving as the reference standard. Sensitivity, specificity,\nPPV, NPV, overall diagnostic accuracy, positive and negative likelihood ratios, and the diagnostic odds ratio\nwere calculated for TVUS and MRI. Corresponding 95% confidence intervals were reported for the diagnostic\nperformance estimates. McNemar’s test was used to compare paired diagnostic classifications between the\ntwo imaging modalities. Interobserver agreement was assessed using Cohen’s kappa coefficient.\nMultivariable logistic regression was performed for the secondary analyses, with model fit indices assessed\nand variance inflation factors examined for multicollinearity. All statistical tests were two-sided, and a P-\nvalue of <0.05 was considered statistically significant.\nEthical considerations\nThe study followed ethical principles applicable to research involving human participants. Before recruiting\nparticipants, the study obtained institutional ethical approval from Rawalpindi Medical College, Rawalpindi,\n \n2026 Ahmed et al. Cureus 18(10): e118898. DOI 10.7759/cureus.118898\n5\n of \n16\n\nPakistan (EC-RMU-331-26). We informed all eligible participants about study procedures, potential risks,\nbenefits, confidentiality, and the voluntary nature of participation, and obtained written informed consent\nbefore enrollment. We coded participant information and treated it confidentially, and used study data only\nfor research purposes.\nReporting standards\nThe study methodology and reporting followed diagnostic-accuracy research standards, with special\nemphasis on participant selection, index-test procedures, reference-standard assessment, blinding, timing,\nand diagnostic performance estimates. The final study flow was documented from eligibility assessment\nthrough completion of TVUS and MRI, laparoscopic surgical assessment, histopathological examination,\nand inclusion in the final analytic cohort.\nResults\nAs shown in Table \n1\n, the study population comprised 350 women with a mean age of 35.67 ± 7.51 years and a\nmean BMI of 25.91 ± 4.06 kg/m². Median gravidity was 2 (IQR, 2), median parity was 1 (IQR, 2), and median\nsymptom duration was 4.7 years (IQR, 4.0). A total of 123 (35.1%) women reported a previous diagnosis of\nendometriosis, and 49 (14.0%) had previously undergone surgery for endometriosis; 41 (11.7%) having had a\nhistory of previous pelvic surgery. Hormonal therapy was reported by 151 (43.1%) women. The most\ncommon presenting symptoms were dysmenorrhea (246, 70.3%), chronic pelvic pain (185, 52.9%), and deep\ndyspareunia (118, 33.7%). Dyschezia was reported by 82 (23.4%) participants, while cyclical bowel symptoms\noccurred in 86 (24.6%). Infertility was present in 74 (21.1%) women. Dysuria and cyclical urinary symptoms\nwere less frequently reported, occurring in 26 (7.4%) and 30 (8.6%) participants, respectively. The cohort was\nrepresentative of a clinical population with a substantial history of endometriosis, and predominantly pain-\ncentered presentations.\nCharacteristic\nOverall (N = 350)\nAge (years)\n35.67 ± 7.51\nBMI (kg/m²)\n25.91 ± 4.06\nGravidity, median (IQR)\n2 (2)\nParity, median (IQR)\n1 (2)\nSymptom duration (years), median (IQR)\n4.7 (4.0)\nPrevious endometriosis diagnosis, n (%)\n123 (35.1)\nPrevious endometriosis surgery, n (%)\n49 (14.0)\nPrevious pelvic surgery, n (%)\n41 (11.7)\nHormonal therapy, n (%)\n151 (43.1)\nDysmenorrhea, n (%)\n246 (70.3)\nChronic pelvic pain, n (%)\n185 (52.9)\nDeep dyspareunia, n (%)\n118 (33.7)\nDyschezia, n (%)\n82 (23.4)\nDysuria, n (%)\n26 (7.4)\nCyclical bowel symptoms, n (%)\n86 (24.6)\nCyclical urinary symptoms, n (%)\n30 (8.6)\nInfertility, n (%)\n74 (21.1)\nTABLE\n 1: \nBaseline demographic and clinical characteristics of the study population (N = 350)\nContinuous variables are presented as mean ± SD or median (IQR), as appropriate; categorical variables are presented as frequency and percentage. No\nvariables included in this table had missing values.\nSD, standard deviation; IQR, interquartile range\n \n2026 Ahmed et al. Cureus 18(10): e118898. DOI 10.7759/cureus.118898\n6\n of \n16\n\nThe overall number of participants with DIE confirmed by histopathology was 213 (60.9%), compared with\n211 (60.3%) at surgery, as shown in Table \n2\n. Anatomically, uterosacral ligament DIE was identified in 102\n(29.1%) cases by histopathology and 104 (29.7%) by surgical evaluation, followed by rectosigmoid DIE in 83\n(23.7%) and 85 (24.3%) cases, respectively. Both evaluations identified rectal DIE in 50 (14.3%) participants,\nand rectovaginal septum involvement in 37 (10.6%) and 38 (10.9%) participants, respectively. Vaginal,\nbladder, and ureteric DIE were identified in 32 (9.1%), 21 (6.0%), and 16 (4.6%) participants, respectively, by\nboth evaluations. Surgical evaluation also revealed obliteration of the pouch of Douglas in 112 (32.0%)\nparticipants. Histopathologically, 205 (58.6%) participants had endometrial glands, 203 (58.0%) had\nendometrial stroma, 174 (49.7%) had fibrosis, and 148 (42.3%) had hemosiderin-laden macrophages.\nFinding\nHistopathology, n (%)\nSurgery, n (%)\nAnatomical distribution of DIE\n-\n-\nOverall DIE\n213 (60.9)\n211 (60.3)\nRectosigmoid DIE\n83 (23.7)\n85 (24.3)\nRectal DIE\n50 (14.3)\n50 (14.3)\nUterosacral ligament DIE\n102 (29.1)\n104 (29.7)\nRectovaginal septum DIE\n37 (10.6)\n38 (10.9)\nVaginal DIE\n32 (9.1)\n32 (9.1)\nBladder DIE\n21 (6.0)\n21 (6.0)\nUreteric DIE\n16 (4.6)\n16 (4.6)\nAdditional surgical finding\n-\n-\nPouch of Douglas obliteration\n-\n112 (32.0)\nHistopathological characteristics\n-\n-\nEndometrial glands present\n205 (58.6)\n-\nEndometrial stroma present\n203 (58.0)\n-\nFibrosis present\n174 (49.7)\n-\nHemosiderin-laden macrophages present\n148 (42.3)\n-\nTABLE\n 2: Histopathological and surgical findings in the study population (N = 350)\nValues are presented as n (%). Histopathology was considered the reference standard. Anatomical site percentages are calculated using the total study\npopulation (N = 350); therefore, categories are not mutually exclusive. Of the 213 participants with histopathologically confirmed DIE, 207 were also\nclassified as having DIE at surgery, whereas six were not identified as having DIE intraoperatively. Conversely, four participants without histopathologically\nconfirmed DIE had a positive surgical assessment. Thus, the surgical assessment identified 211 cases overall, compared with 213 cases confirmed by\nhistopathology.\nDIE, deep infiltrating endometriosis; POD, pouch of Douglas\nTable \n3\n shows that 208 (59.4%) and 205 (58.6%) participants were identified as having DIE on TVUS and\nMRI, respectively. The most common site of DIE was the uterosacral ligament, with 88 (25.1%) cases\nidentified by both modalities, followed by the rectosigmoid region, with 83 (23.7%) cases identified on TVUS\nand 76 (21.7%) on MRI. Rectal DIE was detected in 49 (14.0%) and 48 (13.7%) participants on TVUS and MRI,\nrespectively, while rectovaginal septum involvement was found in 36 (10.3%) and 32 (9.1%), respectively.\nEndometriomas were identified in 68 (19.4%) participants on TVUS and 64 (18.3%) on MRI. Other findings\nincluded pouch of Douglas obliteration (123, 35.1%) and a positive sliding sign (196, 56.0%) on TVUS.\nAdenomyosis was detected in 99 (28.3%) cases on TVUS and 106 (30.3%) on MRI. Quantitative uterine\nmeasurements were similar between the modalities.\n \n2026 Ahmed et al. Cureus 18(10): e118898. DOI 10.7759/cureus.118898\n7\n of \n16\n\nImaging finding\nTVUS, n (%)/mean ± SD\nMRI, n (%)/mean ± SD\nDIE by anatomical site\n-\n-\nRectosigmoid DIE\n83 (23.7)\n76 (21.7)\nRectal DIE\n49 (14.0)\n48 (13.7)\nUterosacral ligament DIE\n88 (25.1)\n88 (25.1)\nRectovaginal septum DIE\n36 (10.3)\n32 (9.1)\nVaginal DIE\n27 (7.7)\n31 (8.9)\nBladder DIE\n21 (6.0)\n20 (5.7)\nUreteric DIE\n13 (3.7)\n12 (3.4)\nOverall DIE\n208 (59.4)\n205 (58.6)\nOther imaging findings\n-\n-\nEndometrioma\n68 (19.4)\n64 (18.3)\nPouch of Douglas obliteration\n123 (35.1)\n-\nPositive sliding sign\n196 (56.0)\n-\nAdenomyosis\n99 (28.3)\n106 (30.3)\nMyometrial abnormality\n62 (17.7)\n62 (17.7)\nQuantitative imaging measurements\n-\n-\nUterine length (mm)\n74.64 ± 9.27\n74.62 ± 9.36\nUterine AP diameter (mm)\n41.95 ± 5.82\n41.98 ± 5.89\nUterine width (mm)\n48.94 ± 6.66\n48.85 ± 6.85\nEndometrial thickness (mm)\n8.60 ± 2.89\n8.64 ± 3.01\nMaximum endometrioma diameter (mm)†\n43.27 ± 13.45\n40.66 ± 14.43\nTABLE\n 3: TVUS and MRI Imaging Findings in the Study Population (N = 350)\nValues are presented as n (%) or mean ± SD. Quantitative measurements were available for all 350 participants except maximum endometrioma diameter,\nwhich was assessed among participants with an endometrioma detected by the respective modality (TVUS, n = 68; MRI, n = 64).\nAP, anteroposterior; DIE, deep infiltrating endometriosis; MRI, magnetic resonance imaging; TVUS, transvaginal ultrasound; SD, standard deviation\nAs shown in Table \n4\n, TVUS demonstrated a sensitivity of 89.2% (95% CI, 84.2-93.0) and a specificity of 86.9%\n(95% CI, 80.0-92.0), with a PPV of 91.3% (95% CI, 86.7-94.8), an NPV of 83.8% (95% CI, 76.7-89.4), and an\noverall accuracy of 88.3% (84.4-91.5). MRI showed slightly higher diagnostic performance, with a sensitivity\nof 91.5% (95% CI, 87.0-94.9), a specificity of 92.7% (95% CI, 87.0-96.4), a PPV of 95.1% (95% CI, 91.2-97.6),\nan NPV of 87.6% (95% CI, 81.1-92.5), and an accuracy of 92.0% (95% CI, 88.6-94.6). MRI also demonstrated a\nhigher LR+ (12.54 vs. 6.79) and a lower LR− (0.091 vs. 0.124), resulting in a substantially higher DOR (137.58\nvs. 54.61). Overall, both modalities had high diagnostic performance, with MRI showing numerically higher\nsensitivity, specificity, and overall accuracy.\n \n2026 Ahmed et al. Cureus 18(10): e118898. DOI 10.7759/cureus.118898\n8\n of \n16\n\nDiagnostic measure\nTVUS\nMRI\nSensitivity, % (95% CI)\n89.2 (84.2-93.0)\n89.2 (84.2-93.0)\nSpecificity, % (95% CI)\n86.9 (80.0-92.0)\n92.7 (87.0-96.4)\nPositive predictive value, % (95% CI)\n91.3 (86.7-94.8)\n95.1 (91.2-97.6)\nNegative predictive value, % (95% CI)\n83.8 (76.7-89.4)\n87.6 (81.1-92.5)\nAccuracy, % (95% CI)\n88.3 (84.4-91.5)\n92.0 (88.6-94.6)\nPositive likelihood ratio, LR+ (95% CI)\n6.79 (4.40-10.47)\n12.54 (6.90-22.81)\nNegative likelihood ratio, LR− (95% CI)\n0.124 (0.084-0.184)\n0.091 (0.058-0.142)\nDiagnostic odds ratio, DOR (95% CI)\n54.61 (28.28-105.45)\n137.58 (61.53-307.64)\nTABLE\n 4: Overall diagnostic performance of TVUS and MRI for detection of DIE\nHistopathology was used as the reference standard. All diagnostic performance estimates are presented with 95% confidence intervals.\nTP, true positive; FP, false positive; FN, false negative; TN, true negative; PPV, positive predictive value; NPV, negative predictive value; LR+, positive\nlikelihood ratio; LR−, negative likelihood ratio; DOR, diagnostic odds ratio; DIE, deep infiltrating endometriosis; MRI, magnetic resonance imaging; TVUS,\ntransvaginal ultrasound\nFigure \n2\n illustrates the overall diagnostic performance of TVUS and MRI for detecting DIE. MRI showed\nnumerically higher sensitivity, specificity, PPV, and NPV than TVUS. However, the overall paired comparison\ndid not demonstrate a statistically significant difference in diagnostic classifications between the two\nmodalities (McNemar’s \nχ\n² = 0.066, P = 0.798). Therefore, the numerical differences should not be interpreted\nas evidence of the statistical superiority of either modality.\nFIGURE\n 2: Overall diagnostic performance of TVUS and MRI for\ndetection of DIE\nDIE, deep infiltrating endometriosis; MRI, magnetic resonance imaging; TVUS, transvaginal ultrasound\nTVUS and MRI were both high-performing modalities across the assessed anatomic sites, as shown in Table\n5\n. For rectosigmoid DIE, MRI showed higher sensitivity and specificity than TVUS (88.0% vs. 86.7% and\n98.9% vs. 95.9%, respectively), with higher accuracy (96.3% vs. 93.7%). Similar patterns were observed for\nrectal, uterosacral ligament, rectovaginal septum, and bladder DIE, where MRI yielded numerically higher\nsensitivity, specificity, and accuracy across several anatomical sites. However, these descriptive differences\nshould be interpreted cautiously because formal paired statistical comparisons were not performed for every\n \n2026 Ahmed et al. Cureus 18(10): e118898. DOI 10.7759/cureus.118898\n9\n of \n16\n\nsite. For vaginal DIE, TVUS had a sensitivity of 75.0%, compared with 78.1% for MRI; PPV was 88.9% for\nTVUS and 80.6% for MRI. For ureteric DIE, TVUS had higher sensitivity (81.3% vs. 68.8%) and perfect\nspecificity (100.0%), although its LR+ and DOR were not estimable because no false-positive cases occurred.\nOverall, both modalities showed excellent diagnostic performance, with MRI generally more specific and\naccurate across most anatomical sites.\nAnatomical\nsite\nModality\nSensitivity, %\n(95% CI)\nSpecificity, %\n(95% CI)\nPPV, %\n(95% CI)\nNPV, %\n(95% CI)\nAccuracy, %\n(95% CI)\nLR+\nLR−\nDOR\nRectosigmoid\nTVUS\n86.7 (77.5-93.2)\n95.9 (92.7-97.9)\n86.7 (77.5-\n93.2)\n95.9 (92.7-\n97.9)\n93.7 (90.6-96.0)\n21.06\n0.138\n152.33\nMRI\n88.0 (79.0-94.1)\n98.9 (96.8-99.8)\n96.1 (88.9-\n99.2)\n96.4 (93.4-\n98.2)\n96.3 (93.7-98.0)\n78.28\n0.122\n642.4\nRectal\nTVUS\n86.0 (73.3-94.2)\n98.0 (95.7-99.3)\n87.8 (75.2-\n95.4)\n97.7 (95.3-\n99.1)\n96.3 (93.7-98.0)\n43\n0.143\n301\nMRI\n90.0 (78.2-96.7)\n99.0 (97.1-99.8)\n93.8 (82.8-\n98.7)\n98.3 (96.2-\n99.5)\n97.7 (95.5-99.0)\n90\n0.101\n891\nUterosacral\nligament\nTVUS\n81.4 (72.4-88.4)\n98.0 (95.4-99.3)\n94.3 (87.2-\n98.1)\n92.7 (88.9-\n95.6)\n90.0 (78.2-96.7)\n40.36\n0.19\n212.31\nMRI\n84.3 (75.8-90.8)\n99.2 (97.1-99.9)\n97.7 (92.0-\n99.7)\n93.9 (90.3-\n96.5)\n94.9 (92.0-96.9)\n104.55\n0.158\n661.13\nRectovaginal\nseptum\nTVUS\n78.4 (61.8-90.2)\n97.8 (95.4-99.1)\n80.6 (64.0-\n91.8)\n97.5 (95.0-\n98.9)\n95.7 (93.0-97.6)\n35.05\n0.221\n158.46\nMRI\n83.8 (68.0-93.8)\n99.7 (98.2-\n100.0)\n96.9 (83.8-\n99.9)\n98.1 (95.9-\n99.3)\n98.0 (95.9-99.2)\n262.24\n0.163\n1612\nVaginal\nTVUS\n75.0 (56.6-88.5)\n99.1 (97.3-99.8)\n88.9 (70.8-\n97.6)\n97.5 (95.2-\n98.9)\n96.9 (94.4-98.4)\n79.5\n0.252\n315\nMRI\n78.1 (60.0-90.7)\n98.1 (95.9-99.3)\n80.6 (62.5-\n92.5)\n97.8 (95.5-\n99.1)\n96.3 (93.7-98.0)\n41.41\n0.223\n185.71\nBladder\nTVUS\n76.2 (52.8-91.8)\n98.5 (96.5-99.5)\n76.2 (52.8-\n91.8)\n98.5 (96.5-\n99.5)\n97.1 (94.8-98.6)\n50.13\n0.242\n207.36\nMRI\n81.0 (58.1-94.6)\n99.1 (97.4-99.8)\n85.0 (62.1-\n96.8)\n98.8 (96.9-\n99.7)\n98.0 (95.9-99.2)\n88.78\n0.192\n461.83\nUreteric\nTVUS\n81.3 (54.4-96.0)\n100.0 (98.9-\n100.0)\n100.0 (75.3-\n100.0)\n99.1 (97.4-\n99.8)\n99.1 (97.5-99.8)\nNE\n0.188\nNE\nMRI\n68.8 (41.3-89.0)\n99.7 (98.3-\n100.0)\n91.7 (61.5-\n99.8)\n98.5 (96.6-\n99.5)\n98.3 (96.3-99.4)\n229.63\n0.313\n732.6\nTABLE\n 5: Site-specific diagnostic performance of TVUS and MRI for detection of DIE\nHistopathology as the reference standard for each anatomical site. Values are based on 350 participants with no missing observations. Confidence\nintervals for sensitivity, specificity, PPV, NPV, and accuracy were calculated using exact binomial methods; likelihood-ratio and DOR confidence intervals\nwere calculated on the log scale. For ureteric TVUS, no false-positive cases were observed; therefore, LR+ or DOR could not be estimated. Site-specific\npercentages are not mutually exclusive. The underlying site-specific crosstabs confirm, for example, rectosigmoid TVUS (72 TP, 11 FP, 11 FN, 256 TN)\nand MRI (73 TP, 3 FP, 10 FN, 264 TN), as well as the corresponding rectal, USL, RVS, vaginal, bladder, and ureteric comparisons.\nCI, confidence interval; DOR, diagnostic odds ratio; DIE, deep infiltrating endometriosis; LR+, positive likelihood ratio; LR−, negative likelihood ratio; MRI,\nmagnetic resonance imaging; NE, not estimable; NPV, negative predictive value; PPV, positive predictive value; TVUS, transvaginal ultrasound\nTable \n6\n shows high agreement between TVUS and MRI in the overall classification of DIE. In 113 (32.3%)\nparticipants, both modalities were negative; in 176 (50.3%) participants, both modalities were positive; and\nin 29 (8.3%) participants, TVUS was negative and MRI was positive, and in 32 (9.1%) participants, TVUS was\npositive and MRI was negative. Overall, TVUS was positive in 208 (59.4%) patients, and MRI was positive in\n205 (58.6%) patients. MRI sensitivity was numerically higher than that of TVUS (91.5% vs. 89.2%), and\nspecificity was also higher (92.7% vs. 86.9%). However, there was no statistically significant difference in\nsensitivity (McNemar’s P = 0.500) or specificity (McNemar’s P = 0.169) between the two methods. The overall\n \n2026 Ahmed et al. Cureus 18(10): e118898. DOI 10.7759/cureus.118898\n10\n of \n16\n\npaired comparison of diagnostic classifications was also not statistically significant (McNemar’s \nχ\n² = 0.066, P\n= 0.798).\nTVUS result\nMRI negative, n (%)\nMRI positive, n (%)\nTotal, n (%)\nTVUS negative\n113 (32.3)\n29 (8.3)\n142 (40.6)\nTVUS positive\n32 (9.1)\n176 (50.3)\n208 (59.4)\nTotal\n145 (41.4)\n205 (58.6)\n350 (100.0)\nTABLE\n 6: Paired comparison of TVUS and MRI for detection of overall DIE\nMcNemar’s test was used for paired comparisons. Sensitivity was compared among participants with histopathology-confirmed DIE, whereas specificity\nwas compared among participants without histopathology-confirmed DIE. The paired difference in sensitivity was not statistically significant (P = .500), nor\nwas the paired difference in specificity (P = .169). For the overall paired classification, McNemar’s \nχ\n² = 0.066, P = 0.798.\nDIE, deep infiltrating endometriosis; MRI, magnetic resonance imaging; TVUS, transvaginal ultrasound\nAs shown in Table \n7\n, interobserver agreement was generally strong to near-perfect for both TVUS and MRI\nacross the assessed anatomical sites. The highest agreement was observed for rectosigmoid DIE, with\nCohen’s \nκ\n values of 0.935 for TVUS and 0.898 for MRI, followed by uterosacral ligament involvement (0.875\nand 0.859, respectively). Agreement remained substantial for rectal, rectovaginal septum, vaginal, and\nbladder involvement, with \nκ\n values ranging from 0.677 to 0.863. Agreement was lowest for ureteric DIE (\nκ\n =\n0.557) in both modalities. All \nκ\n values were statistically significant (P < 0.001), indicating good interobserver\nagreement in the predefined subset of 88 (25.1%) participants.\nAnatomical site\nTVUS Cohen's κ (95% CI)\nMRI Cohen's κ (95% CI)\nRectosigmoid\n0.935 (0.847-1.000)\n0.898 (0.784-1.000)\nRectal\n0.783 (0.603-0.963)\n0.750 (0.540-0.960)\nUterosacral ligament\n0.875 (0.755-0.995)\n0.859 (0.739-0.979)\nRectovaginal septum\n0.823 (0.629-1.000)\n0.727 (0.476-0.978)\nVaginal\n0.863 (0.677-1.000)\n0.677 (0.416-0.938)\nBladder\n0.822 (0.581-1.000)\n0.692 (0.412-0.972)\nUreteric\n0.557 (0.116-0.998)\n0.557 (0.116-0.998)\nTABLE\n 7: Interobserver agreement for TVUS and MRI detection of deep infiltrating endometriosis\nby anatomical site\nWe assessed interobserver agreement in a predefined second-reader subset of 88 participants (25.1% of the study population) using Cohen’s \nκ.\n Values\nare presented as Cohen’s \nκ\n with approximate 95% confidence intervals. Higher \nκ\n values indicate greater interobserver agreement.\nAs indicated in Table \n8\n, a previous diagnosis of endometriosis, a history of prior pelvic surgery, hormonal\ntherapy, dysmenorrhea, chronic pelvic pain, dyschezia, cyclical bowel symptoms, and infertility were\nindependently associated with histopathology-confirmed DIE. Longer symptom duration was associated\nwith lower odds of DIE (aOR, 0.877; 95% CI, 0.805-0.956; P = 0.003), whereas a previous diagnosis of\nendometriosis (aOR, 2.298; 95% CI, 1.309-4.033; P = 0.004), previous pelvic surgery (aOR, 4.563; 95% CI,\n1.650-12.619; P = 0.003), hormonal therapy (aOR, 2.286; 95% CI, 1.342-3.897; P = 0.002), dysmenorrhea\n(aOR, 3.102; 95% CI, 1.757-5.476; P < 0.001), chronic pelvic pain (aOR, 2.438; 95% CI, 1.461-4.066; P =\n0.001), dyschezia (aOR, 2.886; 95% CI, 1.448-5.754; P = 0.003), cyclical bowel symptoms (aOR, 1.923; 95% CI,\n1.027-3.601; P = 0.041), and infertility (aOR, 2.342; 95% CI, 1.200-4.569; P = 0.013) were associated with\nhigher odds of DIE. Age, BMI, previous endometriosis surgery, deep dyspareunia, dysuria, and cyclical\nurinary symptoms were not independently associated with histopathology-confirmed DIE in the\nmultivariable model. The overall model was significant (\nχ\n² = 108.943, P < 0.001) and correctly classified\n74.9% of cases.\n \n2026 Ahmed et al. Cureus 18(10): e118898. DOI 10.7759/cureus.118898\n11\n of \n16\n\nPredictor\nAdjusted OR\n95% CI\nP-value\nAge, per year\n0.974\n0.942-1.006\n0.131\nBMI, per kg/m²\n0.989\n0.928-1.052\n0.742\nSymptom duration, per year\n0.877\n0.805-0.956\n0.003\nPrevious endometriosis diagnosis\n2.298\n0.805-0.956\n0.004\nPrevious endometriosis surgery\n2.278\n0.988-5.251\n0.053\nPrevious pelvic surgery\n4.563\n1.650-12.619\n0.003\nHormonal therapy\n2.286\n1.342-3.897\n0.002\nDysmenorrhea\n3.102\n1.757-5.476\n<0.001\nChronic pelvic pain\n2.438\n1.461-4.066\n0.001\nDeep dyspareunia\n1.615\n0.941-2.773\n0.083\nDyschezia\n2.886\n1.448-5.754\n0.003\nDysuria\n1.859\n0.619-5.584\n0.268\nCyclical bowel symptoms\n1.923\n1.027-3.601\n0.041\nCyclical urinary symptoms\n1.815\n0.672-4.904\n0.24\nInfertility\n2.342\n1.200-4.569\n0.013\nTABLE\n 8: Multivariable analysis of clinical factors associated with histopathology-confirmed DIE\nModel statistics: likelihood-ratio \nχ\n²(15) = 108.943, P < 0.001; −2 log likelihood = 359.625; Cox and Snell R² = 0.267; Nagelkerke R² = 0.363; Hosmer-\nLemeshow \nχ\n²(8) = 12.113, P = 0.146; overall classification accuracy = 74.9%. We derived adjusted ORs from multivariable binary logistic regression. The\noutcome was histopathology-confirmed DIE (0 = no, 1 = yes). We entered continuous variables as continuous predictors; binary clinical variables were\ncoded 0 = no and 1 = yes, with absence/no as the reference category. OR > 1 indicates higher odds of histopathology-confirmed DIE, whereas OR < 1\nindicates lower odds.\nCI, confidence interval; DIE, deep infiltrating endometriosis; OR, odds ratio\nDiscussion\nPrincipal findings and overall diagnostic performance\nThis prospective paired study on diagnostic accuracy found that both TVUS and MRI were highly accurate in\ndiagnosing DIE compared with histopathology. TVUS sensitivity, specificity, and overall accuracy were\n89.2%, 86.9%, and 88.3%, respectively, while MRI values were 91.5%, 92.7%, and 92.0%, respectively. MRI\nalso showed a modest numerical advantage in PPV, NPV, positive likelihood ratio, and diagnostic odds ratio,\nindicating numerically higher diagnostic performance for MRI across several measures in this cohort. These\nresults align with past studies showing that both TVUS and MRI have high diagnostic accuracy for DIE. A\nsystematic review and meta-analysis of 30 studies with 4,565 subjects showed pooled sensitivities of 76%\nand 82% for TVUS and MRI, respectively, and specificities of 94% for TVUS and 87% for MRI, with high\ndiagnostic accuracy for both modalities \n[13]\n. Likewise, a prospective paired comparison study of TVUS with\ncontrast-enhanced MR colonography demonstrated good diagnostic accuracy for both imaging modalities\n(TVUS, 89.2%; MRI, 87.2%) \n[14]\n. The results suggest that both modalities can serve as reliable preoperative\ntools for detecting DIE \n[13,15]\n and that MRI had a slight numerical advantage in overall diagnostic\nperformance in this study. Although MRI demonstrated numerically higher values for several diagnostic\nperformance measures, the overall paired comparison did not show a statistically significant difference\nbetween TVUS and MRI classifications (McNemar’s \nχ\n² = 0.066, P = 0.798). Therefore, these findings do not\nestablish the superiority of MRI over TVUS, and the numerical differences should be interpreted cautiously.\nDiagnostic performance of standardized TVUS\nTVUS’s high diagnostic accuracy is important, as it demonstrates the added value of a standardized, expertly\nperformed ultrasound examination in the perioperative evaluation of DIE \n[16]\n. In this study, TVUS was\ncarried out using a systematic IDEA-based approach, including evaluation of the uterosacral ligaments,\nrectovaginal septum, vagina, rectum, rectosigmoid, bladder, ureters, pouch of Douglas, and pelvic organ\nmobility. This extensive evaluation may have helped achieve the improved diagnostic performance\n \n2026 Ahmed et al. Cureus 18(10): e118898. DOI 10.7759/cureus.118898\n12\n of \n16\n\nobserved, as recent validation studies show increasing sensitivity and overall accuracy as the IDEA protocol\nsteps were added \n[17]\n. Current approaches to endometriosis imaging emphasize systematic anatomical\nassessment and examination by adequately trained operators rather than relying on nonspecific pelvic\nultrasonography alone \n[4]\n.\nDiagnostic performance and clinical role of MRI\nMRI demonstrated a modest numerical advantage in diagnostic performance in the present cohort,\nparticularly for specificity and positive likelihood ratio. The higher specificity (92.7%) and LR+ (12.54)\nindicate a strong ability to exclude disease and significantly increase the probability of DIE with MRI\npositivity. This aligns with previous head-to-head evidence that MRI can be more specific and have higher\nlikelihood ratios than TVUS in some anatomical locations. In a systematic review and meta-analysis of\npaired studies, MRI showed higher specificity (93% vs. 86%) and LR+ (10.4 vs. 4.8) than TVUS for uterosacral\nligament DIE. However, the diagnostic performance of the two modalities was broadly similar for\nrectosigmoid and rectovaginal septal disease \n[14]\n. This could be especially useful in extensive or more\ncomplex disease, where multiplanar MRI and compartment-based evaluation allow thorough mapping and\nprovide additional data for preoperative surgical planning \n[18]\n. The standardized MRI protocol used in this\nstudy incorporated multiplanar T2-weighted sequences, T1-weighted/fat-suppressed imaging, and\nsystematic evaluation of multiple pelvic compartments, consistent with established ESUR recommendations\n[12]\n. An experienced pelvic radiologist interpreted all MRI examinations.\nSite-specific diagnostic performance\nSite-specific findings further show that diagnostic performance varied by anatomical compartment. MRI\nshowed a numerical advantage in accuracy for several compartments in the present cohort, including\nrectosigmoid, rectal, uterosacral ligament, rectovaginal septal, and bladder DIE. For instance, MRI had\n96.3% accuracy versus 93.7% for rectosigmoid disease and 97.7% versus 96.3% for rectal disease. MRI\ndiagnosed rectovaginal septal disease with 98.0% accuracy, whereas TVUS achieved 95.7%. These results\nsuggest that MRI may add value when more precise delineation of disease distribution is needed.\nNevertheless, the advantage was not universal. TVUS was more sensitive than MRI for ureteric DIE (81.3% vs.\n68.8%) and had perfect specificity (100.0%). This is consistent with recent comparative data showing that\nthe diagnostic performance of MRI and TVUS is site-dependent. A recent meta-analysis of rectosigmoid DIE\nidentified that both MRI and TVUS had high diagnostic accuracy and were comparable with respect to\nspecificity and sensitivity, with slightly higher specificity for MRI than for TVUS (95% vs. 92%), but with no\nsignificant difference between them in terms of specificity or sensitivity \n[19]\n. Likewise, a recent systematic\nreview found that TVUS and MRI are complementary procedures, with TVUS being highly accurate for\nposterior-compartment disease and MRI providing greater body coverage and higher sensitivity for anterior\nand multicompartmental lesions \n[9]\n. Thus, the choice of imaging modality should consider the suspected\nanatomical distribution rather than assuming that one modality is uniformly superior. However, because\nformal paired statistical comparisons were not performed for each anatomical site, the observed differences\nshould be interpreted descriptively and should not be considered evidence of statistically significant\nsuperiority of either modality at individual sites.\nPaired comparison and complementary role of TVUS and MRI\nThe paired analysis showed that both modalities classified 113 (32.3%) participants as negative and 176\n(50.3%) as positive. Discordant classifications occurred in 29 (8.3%) participants who were TVUS-\nnegative/MRI-positive and 32 (9.1%) who were TVUS-positive/MRI-negative. The lack of a statistically\nsignificant difference in the paired classifications indicates that this cohort did not show a statistically\nsignificant difference in marginal paired classifications. Earlier comparative analyses support this view,\nshowing similar diagnostic accuracy for TVUS and MRI across different anatomic sites, with MRI superior at\nsome sites and TVUS at others \n[15,20]\n. TVUS may serve as an effective first-line imaging and mapping\nexamination when performed by appropriately trained operators. At the same time, MRI can provide\ncomplementary anatomical characterization when more extensive or anatomically complex disease requires\nfurther delineation \n[4]\n.\nInterobserver agreement and reproducibility\nThe reproducibility analysis also confirms the reliability of the imaging results. Interobserver agreement was\nhigh for most anatomical sites, with the highest agreement for rectosigmoid DIE (0.935 for TVUS and 0.898\nfor MRI) and uterosacral ligament disease (0.875 for TVUS and 0.859 for MRI). Agreement was lower for\nureteric disease (\nκ\n = 0.557 for both modalities), indicating that urinary tract disease might be harder to\ndiagnose consistently in this cohort. These results support previous work showing good-to-high\ninterobserver agreement between TVUS and MRI in diagnosing deep endometriosis and suggest site- and\nexaminer-dependent variations in reproducibility \n[21,22]\n. The results thus highlight the need for specialist\ntraining and standardization of imaging assessment and reporting, especially for less common or more\ncomplex locations.\nClinical predictors of histopathology-confirmed DIE\n \n2026 Ahmed et al. Cureus 18(10): e118898. DOI 10.7759/cureus.118898\n13\n of \n16\n\nThe secondary clinical analysis identified several factors independently associated with histopathologically\nconfirmed DIE, including a previous diagnosis of endometriosis, previous pelvic surgery, hormonal therapy,\ndysmenorrhea, chronic pelvic pain, dyschezia, cyclical bowel symptoms, and infertility. Previous pelvic\nsurgery (aOR 4.563) and dyschezia (aOR 2.886) were among the clinical factors independently associated\nwith histopathology-confirmed DIE. Dysmenorrhea showed the strongest association among the symptom\nvariables examined (aOR 3.102). These results corroborate previous evidence showing that deep\nendometriosis was associated with severe dysmenorrhea, dyschezia, and infertility, and that the association\nwas stronger when more than one pelvic compartment was affected \n[23]\n. Detailed clinical assessment has\nalso been shown to predict DIE in the preoperative setting, with combinations of clinical symptoms helping\nmake this prediction \n[24]\n. However, clinical symptoms should be considered supporting predictors rather\nthan substitutes for anatomical imaging, and MRI can provide complementary information about the\nlocation and extent of deep disease and can confirm the diagnosis of clinically suspected DIE when\nappropriate diagnostic criteria are used \n[25]\n.\nStrengths of the study\nA major strength of this study was its prospective paired design, in which all 350 participants underwent\nboth imaging modalities before laparoscopic assessment and histopathological evaluation. The within-\nparticipant design reduced variability across patient populations in the different imaging groups and allowed\ndirect comparisons between TVUS and MRI. Consecutive recruitment, predefined imaging protocols, blinded\ninterpretation between modalities, systematic anatomical assessment, and histopathology as the reference\nstandard further strengthened the study’s methodological rigor.\nLimitations\nThis study had several limitations. The relatively high prevalence of DIE in the participants may limit\ngeneralizability to lower-risk or primary care populations since participants were already scheduled for\nlaparoscopic evaluation. Also, no formal a priori sample size calculation was performed, as the final cohort\ncomprised consecutive eligible participants recruited during the study period. Consequently, the precision of\nsome diagnostic-accuracy estimates, particularly for less frequently affected anatomical sites such as the\nureters and bladder, may be limited by the relatively small number of histopathologically confirmed cases.\nThe findings for these sites should therefore be interpreted cautiously. Second, experts conducted the\nexaminations in a standardized fashion, and diagnostic performance may be higher in more specialized\nsettings. Third, histopathology was used as the reference standard; however, histopathological findings are\nlimited to the tissue sampled and may not fully capture the overall anatomical extent of disease across the\npelvis. Consequently, lesions in unsampled areas may not be represented in the histopathological\nassessment, potentially affecting site-specific diagnostic comparisons. Fourth, interobserver agreement was\nexamined in only 88 participants, which reduced the generalizability of reproducibility estimates. Lastly, the\nstudy did not assess downstream clinical outcomes. Therefore, it cannot be concluded whether the absolute\ndifferences between MRI and TVUS lead to better surgical planning, fewer surgical complications, more\ncost-effective care, or better patient outcomes.\nConclusions\nIn this prospective paired diagnostic-accuracy study, both standardized TVUS and MRI had high diagnostic\nperformance when compared with the study reference standard for DIE. Although MRI showed numerically\nhigher diagnostic performance for several anatomical sites, the paired analysis revealed no statistically\nsignificant difference between the modalities in overall positive/negative classification. MRI showed\nnumerically higher sensitivity, specificity, and overall accuracy. The results confirm that expert,\nstandardized TVUS is an appropriate first-line imaging examination, and MRI can provide additional\nanatomical data for women with large, complex, or indeterminate disease and those who need detailed\npreoperative mapping. The results should be interpreted in the context of the specialist referral population\nand the limitations of tissue-based reference-standard assessment.\nAdditional Information\nAuthor Contributions\nAll authors have reviewed the final version to be published and agreed to be accountable for all aspects of the\nwork.\nConcept and design:\n  \nSaadia Hassan, Amina \nIqbal, Urooj Anwar, Nida \nWaqas, Titilayo \nOlawunmi ,\nGyanendra KC, Rachel Marsha Manoj Stanley\nAcquisition, analysis, or interpretation of data:\n  \nSaadia Hassan, Doaa Ahmed, Amina \nIqbal, Gauri\nSreeram, Urooj Anwar, Nida \nWaqas, Gyanendra KC, Vanessa Colaco, Rachel Marsha Manoj Stanley, Nazia\nSumbul Mahir\nDrafting of the manuscript:\n  \nSaadia Hassan, Doaa Ahmed, Amina \nIqbal, Gauri \nSreeram, Urooj Anwar, Nida\nWaqas, Titilayo \nOlawunmi , Gyanendra KC, Vanessa Colaco, Rachel Marsha Manoj Stanley, Nazia Sumbul\n \n2026 Ahmed et al. Cureus 18(10): e118898. DOI 10.7759/cureus.118898\n14\n of \n16\n\nMahir\nCritical review of the manuscript for important intellectual content:\n  \nDoaa Ahmed, Gauri \nSreeram,\nTitilayo \nOlawunmi , Vanessa Colaco, Nazia Sumbul Mahir\nDisclosures\nHuman subjects:\n Informed consent for treatment and open access publication was obtained or waived by all\nparticipants in this study. Rawalpindi Medical College, Rawalpindi, Pakistan issued approval EC-RMU-331-\n26. \nAnimal subjects:\n All authors have confirmed that this study did not involve animal subjects or tissue.\nConflicts of interest:\n In compliance with the ICMJE uniform disclosure form, all authors declare the\nfollowing: \nPayment/services info:\n All authors have declared that no financial support was received from\nany organization for the submitted work. \nFinancial relationships:\n All authors have declared that they have\nno financial relationships at present or within the previous three years with any organizations that might\nhave an interest in the submitted work. \nOther relationships:\n All authors have declared that there are no\nother relationships or activities that could appear to have influenced the submitted work.\nAcknowledgements\nDoaa Ahmed and Gauri Sreeram contributed equally to this manuscript and should be considered co-first\nauthors.\nReferences\n1\n. \nBecker CM, Bokor A, Heikinheimo O, et al.: \nESHRE guideline: endometriosis\n. Hum Reprod Open. 2022,\n2022:hoac009. \n10.1093/hropen/hoac009\n2\n. \nThomassin-Naggara I, Dolciami M, Chamie LP, et al.: \nESUR consensus MRI for endometriosis: indications,\nreporting, and classifications\n. Eur Radiol. 2025, 35:7260-8. \n10.1007/s00330-025-11579-0\n3\n. \nAbike F, Tanoglu FB, Sidar G: \nDeep pelvic endometriosis: clinical features, diagnosis, and treatment - a\ncomprehensive review\n. Arch Gynecol Obstet. 2025, 312:1857-69. \n10.1007/s00404-025-08187-0\n4\n. \nCondous G, Gerges B, Thomassin-Naggara I, et al.: \nNon-invasive imaging techniques for diagnosis of pelvic\ndeep endometriosis and endometriosis classification systems: an International Consensus Statement\n. Eur J\nRadiol. 2024, 176:111450. \n10.1016/j.ejrad.2024.111450\n5\n. \nSingh SS, Allaire C, Al-Nourhji O, et al.: \nGuideline No. 449: diagnosis and impact of endometriosis - a\nCanadian guideline\n. J Obstet Gynaecol Can. 2024, 46:102450. \n10.1016/j.jogc.2024.102450\n6\n. \nDaniilidis A, Grigoriadis G, Dalakoura D, D'Alterio MN, Angioni S, Roman H: \nTransvaginal ultrasound in the\ndiagnosis and assessment of endometriosis-an overview: how, why, and when\n. Diagnostics (Basel). 2022,\n12:2912. \n10.3390/diagnostics12122912\n7\n. \nTang X, Ling R, Gong J, et al.: \nDeep infiltrating endometriosis MR imaging with surgical correlation\n. Quant\nImaging Med Surg. 2018, 8:187-95. \n10.21037/qims.2018.01.10\n8\n. \nVanBuren W, Feldman M, Shenoy-Bhangle AS, et al.: \nRadiology state-of-the-art review: endometriosis\nimaging interpretation and reporting\n. Radiology. 2024, 312:e233482. \n10.1148/radiol.233482\n9\n. \nO'Leary M, Neary C, Lawrence E: \nThe diagnostic accuracy of magnetic resonance imaging versus\ntransvaginal ultrasound in deep infiltrating endometriosis and their impact on surgical decision-making: a\nsystematic review\n. Diagnostics (Basel). 2025, 15:2856. \n10.3390/diagnostics15222856\n10\n. \nTong A, Cope AG, Waters TL, McDonald JS, VanBuren WM: \nBest practices: ultrasound versus MRI in the\nassessment of pelvic endometriosis\n. AJR Am J Roentgenol. 2024, 223:e2431085. \n10.2214/AJR.24.31085\n11\n. \nIndrielle-Kelly T, Frühauf F, Fanta M, et al.: \nDiagnostic accuracy of ultrasound and MRI in the mapping of\ndeep pelvic endometriosis using the International Deep Endometriosis Analysis (IDEA) consensus\n. Biomed\nRes Int. 2020, 2020:3583989. \n10.1155/2020/3583989\n12\n. \nBazot M, Bharwani N, Huchon C, et al.: \nEuropean society of urogenital radiology (ESUR) guidelines: MR\nimaging of pelvic endometriosis\n. Eur Radiol. 2017, 27:2765-75. \n10.1007/s00330-016-4673-z\n13\n. \nZhang X, He T, Shen W: \nComparison of physical examination, ultrasound techniques and magnetic\nresonance imaging for the diagnosis of deep infiltrating endometriosis: a systematic review and meta-\nanalysis of diagnostic accuracy studies\n. Exp Ther Med. 2020, 20:3208-20. \n10.3892/etm.2020.9043\n14\n. \nVimercati A, Achilarre MT, Scardapane A, et al.: \nAccuracy of transvaginal sonography and contrast-\nenhanced magnetic resonance-colonography for the presurgical staging of deep infiltrating endometriosis\n.\nUltrasound Obstet Gynecol. 2012, 40:592-603. \n10.1002/uog.11179\n15\n. \nGuerriero S, Saba L, Pascual MA, Ajossa S, Rodriguez I, Mais V, Alcazar JL: \nTransvaginal ultrasound vs\nmagnetic resonance imaging for diagnosing deep infiltrating endometriosis: systematic review and meta-\nanalysis\n. Ultrasound Obstet Gynecol. 2018, 51:586-95. \n10.1002/uog.18961\n16\n. \nLeonardi M, Uzuner C, Mestdagh W, et al.: \nDiagnostic accuracy of transvaginal ultrasound for detection of\nendometriosis using International Deep Endometriosis Analysis (IDEA) approach: prospective international\npilot study\n. Ultrasound Obstet Gynecol. 2022, 60:404-13. \n10.1002/uog.24936\n17\n. \nMcClenahan P, Mak J, Uzuner C, Espada Vaquero M, Reid S, Eathorne A, Condous G: \nDiagnostic test\naccuracy of each of the four IDEA ultrasound steps in predicting presence of deep endometriosis\n.\nUltrasound Obstet Gynecol. 2026, 67:679-83. \n10.1002/uog.70208\n18\n. \nThomassin-Naggara I, Dolciami M, Chamie LP, et al.: \nESUR consensus MRI for endometriosis: protocol,\nlexicon, and compartment-based analysis\n. Eur Radiol. 2025, 35:7272-86. \n10.1007/s00330-025-11611-3\n19\n. \nXu Z, Li Y, Wang Y, Wan Y, Chen J: \nTransvaginal ultrasound and magnetic resonance imaging in detecting\nrectosigmoid deep infiltrating endometriosis: a comparative meta-analysis\n. Front Med (Lausanne). 2025,\n12:1552185. \n10.3389/fmed.2025.1552185\n20\n. \nGuerriero S, Alcázar JL, Pascual MA, et al.: \nDeep infiltrating endometriosis: comparison between 2-\n \n2026 Ahmed et al. Cureus 18(10): e118898. DOI 10.7759/cureus.118898\n15\n of \n16\n\ndimensional ultrasonography (US), 3-dimensional US, and magnetic resonance imaging\n. J Ultrasound Med.\n2018, 37:1511-21. \n10.1002/jum.14496\n21\n. \nGuerriero S, Pascual M, Ajossa S, et al.: \nThe reproducibility of ultrasonographic findings of rectosigmoid\nendometriosis among examiners with different level of expertise\n. J Ultrasound Med. 2022, 41:403-8.\n10.1002/jum.15717\n22\n. \nHarth S, Kaya HE, Zeppernick F, et al.: \nApplication of the #Enzian classification for endometriosis on MRI:\nprospective evaluation of inter- and intraobserver agreement\n. Front Med (Lausanne). 2023, 10:1303593.\n10.3389/fmed.2023.1303593\n23\n. \nRocha TP, Andres MP, Carmona F, Baracat EC, Abrão MS: \nDeep endometriosis: the involvement of multiple\npelvic compartments is associated with more severe pain symptoms and infertility\n. Reprod Sci. 2023,\n30:1668-75. \n10.1007/s43032-022-01104-9\n24\n. \nStratton P: \nThe association of clinical symptoms with deep infiltrating endometriosis: the importance of\nthe preoperative clinical assessment\n. Hum Reprod. 2014, 29:1627-8. \n10.1093/humrep/deu131\n25\n. \nHarth S, Roller FC, Zeppernick F, Meinhold-Heerlein I, Krombach GA: \nDeep infiltrating endometriosis:\ndiagnostic accuracy of preoperative magnetic resonance imaging with respect to morphological criteria\n.\nDiagnostics (Basel). 2023, 13:1794. \n10.3390/diagnostics13101794\n \n2026 Ahmed et al. Cureus 18(10): e118898. DOI 10.7759/cureus.118898\n16\n of \n16","source_license":"CC0","license_restricted":false}