{"paper_id":"89393e80-18bc-45a6-8aa0-320cfe7ca9ab","body_text":"International Journal of Radiation Research, October 2024 Volume 22, No 4 \nLogistic regression analysis of imaging characteristics of \ntransvaginal ultrasonography for predicting severe \nendometriosis by r-ASRM classification via laparoscopy \nINTRODUCTION \nEndometriosis (EMT) refers to the presence of \nactive EMT tissues (glands and stroma) outside of the \nuterus (1), causing unbearable chronic pain and               \ninfertility. Approximately 30% ~ 50% of women with \nendometriosis struggle with infertility, and 25% ~ \n50% of women with infertility also have                       \nendometriosis (2). Laparoscopic surgery is the most \ncommon intervention for EMT (3). Since 1996, the \nrevised American Society for Reproductive Medicine \n(r-ASRM) classification system has been widely used \nworldwide for the staging of EMT lesions via                  \nlaparoscopic visualization, categorizing EMT in the \npelvic area into four stages: minimal (stage I), mild \n(stage II), moderate (Stage III), and severe (stage IV) \nEMT (4). However, r -ASRM classification is done after \nlaparoscopic surgery, which also indicates that the \nnecessity of laparoscopic surgery should be                   \neffectively evaluated. Since the r -ASRM classification \nsystem does not include many affected organs and \nanatomical structures in the pelvic cavity, it has not \nbeen directly applied in preoperative examination (5). \nThe early diagnosis of EMT before treatment and  \naccurate assessment after treatment is essential for \nthe effective clinical management (6,7). \nThe diagnosis of EMT is challenged due to the    \nheterogeneity of the disease, uncertainty in                 \npathogenesis, asymptom, and complication with           \nadenomyosis (8). Currently, the diagnosis of EMT     \nQ. Su1, H. Luo2, J. Guo1, H. Ning2, Zh. Xu2, Ch. Zhen3, J. Chen2, F. Wang3,          \nQ. Li4, P. Wang1* \n \n1Department of Ultrasonography, The Third Affiliated Hospital, Southern Medical University, Guangzhou, 510630, \nGuangdong, China \n2Department of Ultrasound, Foshan Women and Children Hospital Affiliated to Guangdong Medical University, \n528000, Guangdong, China \n3Department of Ultrasound, The First People's Hospital of Foshan, Foshan,528000, Guangdong, China  \n4Department of Gynecology, Foshan Women and Children Hospital Affiliated to Guangdong Medical University, \nFoshan, 528000, Guangdong, China \nABSTRACT \nBackground: To explore the relation between results of transvaginal ultrasonography \nand the revised American Society for Reproductive Medicine (r -ASRM) staging based \non laparoscopy in patients with endometriosis (EMT) and to establish a prediction \nmodel for risk of severe endometriosis based on the imaging characteristics of \ntransvaginal ultrasonography. Materials and Methods: A retrospective study was \nperformed between April 2022 and May 2023 on women with EMT. The laparoscopic \nsurgery results were used as the golden standard. Patients were divided into the \nminimal-to-moderate endometriosis (stage I -III) and severe endometriosis (stage IV) \ngroups based on r -ASRM classification. The transvaginal ultrasonography imaging \ncharacteristics were extracted to establish a logistic regression model. Results: Among \n200 patients with endometriosis, there were 78 cases of minimal -to-moderate \nendometriosis (stage I -III) and 122 cases of severe endometriosis (stage IV). \nMultivariate analysis showed that the maximum diameter of endometriomas in the \nright ovary, occurrence of unilateral or bilateral ovarian endometriomas, and degree \nof obliteration of the rectouterine pouch were independent predictors for the r -ASRM \nstage of endometriosis. The logistic regression model established using the above \nthree variables had a sensitivity of 82.0%, a specificity of 93.6%, an accuracy of 86.5%, \nand an area under the curve of 0.933 (standard error 0.016, P < 0.005, 95% confidence \ninterval: 0.901, 0.965). Conclusion: Based on laparoscopic visualization, the radiomic \nfeatures of preoperative transvaginal ultrasonography in patients with endometriosis \nwere correlated with the endometriotic stage. The established model using these \ncharacteristics accurately predicted the r -ASRM stage of endometriosis after \nlaparoscopic surgery.  \n►  Original article \nKeywords: Endometriosis, laparoscope, \nregression analysis, ultrasonography.  \n*Corresponding author: \nPing Wang, Ph.D., \nE-mail: \nnysycskwp123@smu.edu.cn  \nReceived: January 2023  \nFinal revised: January 2024 \nAccepted: February 2024  \nInt. J. Radiat. Res., October 2024;         \n22(4): 991-998 \nDOI: 10.61186/ijrr.22.4.998 \n\nincludes ultrasonography, magnetic resonance              \nimaging (MRI), measurement of the serum cancer \nantigen 125 level, and laparoscopic surgery.                   \nHistological confirmation of ectopic endometrial          \ntissue via laparoscopy remains the gold standard for \nthe diagnosis of EMT, while the applicability is          \nlimited by the invasive procedure (6). Transvaginal \nultrasonography has the advantages as follows: (i) \nthe accurate prediction of EMT severity by displaying \nthe fine structure of organs and tissues in the pelvic \ncavity; (ii) the effective evaluation of the distribution \nand infiltration of deep infiltrating EMT (DIE) lesions \nin various parts of the pelvic cavity. Transvaginal  \nultrasonography has become the first -line approach \nfor screening EMT (9,10). Compared with MRI,                    \ntransvaginal ultrasonography shows relatively higher \nspecificity in the detection of deep infiltrating              \nendometriosis (11). Moreover, the transvaginal               \nultrasonography and MRI show no systemic                     \ndifference in the EMT detection compared with the \nintraoperative measurement, and transvaginal               \nultrasonography is more recommended for the              \ndiagnostic examination due to its high availability, \nlow cost and similar accuracy relative to MRI (12,13). \nIn the present study, we analyzed the correlation \nbetween the transvaginal ultrasonography imaging \ncharacteristics in EMT and r-ASRM classification after \nlaparoscopic surgery to assess the surgical difficulty \npreoperatively, which might provide the theoretical \nbasis and novel clues for the effective and accurate \ndiagnosis and management of EMT. \n \n \nMATERIALS AND METHODS \n \nResearch participants \nWe retrospectively analyzed the clinical data of \n200 patients (aged 19-51 years)  with DIE screened \nby transvaginal ultrasonography in Foshan      \nMaternal and Child Health  Hospital (Guangdong \nProvince, China) from August 2021 to December \n2022.  All patients underwent laparoscopic surgery \nfor treatment, and EMT was pathologically                   \nconfirmed. The clinical data collected in this study \nincluded clinical symptoms  and signs, surgical              \nrecords, surgical staging, and pathological data. The \nlaparoscopic surgery-based r-ASRM classification \nincluded four EMT stages: minimal (stage I), mild \n(stage II), moderate (stage III), and severe (stage \nIV). Laparoscopic surgery  results were used as the \ngold standard. Patients were divided into two \ngroups based on their r-ASRM stage: the minimal-to-\nmoderate group (stage I-III) and the severe group \n(stage IV). \n \nResearch apparatus and methods \nA Samsung WS80 A Ultrasound Machine was \nused, with an intracavity probe frequency  ranging \nfrom 5 to 9 Mega Hertz ( MHz) and 9 to 12 MHz. An \n992 \nabdominal probe was used with a frequency ranging \nfrom 5 to 9 MHz. \n \nUltrasonography \nThe patient took the lithotomy position , after \nwhich their vagina,  uterus, anterior pelvic cavity, \nposterior pelvic cavity, and specific tender points \nwere successively scanned through the vagina. The \ncollected cases were independently reviewed by two \nblinded radiologists (each with more than 5 years of \nultrasonography experience ). They analyzed the              \nimages following  the consensus issued by the                 \nInternational Deep EMT Analysis Group (referred to \nas consensus) (14) and discrepancies were resolved by \ndiscussion between  the two radiologists. The            \nlocation, number, size and imaging manifestation of  \nEMT lesions were recorded in a standardized report. \n \nLaparoscopic surgery \nExperienced surgeons operated on all patients. \nThe pelvic and abdominal cavity and the EMT lesions \nwere assessed. After removing the connective tissues \nto expose the affected areas completely, the lesions, \nincluding the adjacent  tissue up to 0.5 centimeters \nfrom the outer edge, were excised. The appearance of \nthe abdominal adhesions, the uterus, the uterosacral \nligament, fallopian tubes, the rectum and the                \npresence of EMT lesions were recorded. A                      \npostoperative r- ASRM scoring table was completed \nto classify the EMT as minimal (stage I), mild (stage \nII), moderate (stage III), or severe (stage IV). \n \nStatistical analysis \nThe SPSS23.0 software package (IBM, Armonk, \nNY, USA) was used for the statistical analysis  of the \ndata. Normally distributed data were compared \nusing the t- test and presented  as the  mean ±              \nstandard deviation. Non-normally distributed data \nwere compared  using the rank sum test and are          \npresented as median (M) and interquartile range \n(P25, P75). Count data were compared using the          \nchi-square test and were presented as frequency (n) \nand percentage (%). Multivariate analysis was               \ncarried out by binary logistic regression. P < 0.05 was \nconsidered significant. \n \n \nRESULTS \n \nResults of r-ASRM classification of endometriotic \nlesions \nAmong 200 patients with EMT, according to the            \nr-ASRM classification, there were 78 cases of minimal\n-to-moderate EMT (stage I – III) and 122 cases of     \nsevere EMT (stage IV). Table 1 shows the clinical \ncharacteristics and affected sites of all patients in this \nstudy. The ultrasound imaging features of different \nlesions are shown in figures 1 -4. As shown in figure 1, \nthe ovaries of this patient were partly joined            \nInt. J. Radiat. Res., Vol. 22 No. 4, October 2024 \n\ntogether, termed as “kissing ovaries” sign, and             \nbilateral ovarian endometriotic cysts were also           \nobserved. Figure 2 showed that this DIE patient         \npresented nodules infiltrating the right uterosacral \nligament. Figure 3 exhibited that the lesion of this DIE \npatient was located at the intestines with blur and \nspiculate boundary as the “Indian headdress”. Figure \n4 showed the rectouterine pouch obliteration. The \nlesion adheres to surrounding tissues, and the rectum \nslid against the uterine wall, termed as uterine sliding \nsign. Bilateral uterosacral ligament and intestinal DIE \nunder the hysteroscopic are shown in figures 5 and 6. \nAs shown in figure 5, the observation under a           \nhysteroscope identified the infiltrating nodules at \nbilateral uterosacral ligament of the DIE patient.         \nFigure 6 indicated that the patient presented          \nendometriotic cyst in the right ovarian and the          \nintestinal lesion under a hysteroscope. \nSu et al. / A model using transvaginal ultrasonography to predict EMT  993 \n  I-III (n=78) IV (n=122) t/c2 P \nAge (years) 33.62±5.87 34.33±7.61 0.742 0.459 \nAge of menarche 12.87±1.21 13.01±1.08 0.832 0.407 \nmarital status     0.619 0.431 \nMarried (n) 59 98     \nUnmarried (n) 19 24     \nPregnancy history     1.518 0.218 \nYes 24 48     \nNo 54 74     \nBMI (kg/m2) 23.8±4.2 24.1±3.9 0.507 0.613 \nInfertility 22 41 0.643 0.423 \nUterine adenomyosis         \nNo 61 69 9.801 0.002 \nYes 17 53     \nUterine size         \nNo 68 89 5.707 0.017 \nYes 10 33     \nOvarian lesions         \nNo 9 3 57.199 0.000 \nUnilateral involvement 65 49     \nBilateral involvement 4 70     \nOvarian dislocation movement         \nExist 25 11 38.359 0.000 \nOne-sided disappearance 34 28     \nBilateral disappearance 19 83     \nRectal notch with or without occlusion         \nNo 42 5 82.715 0.000 \nOne-sided occlusion 32 51     \nBilateral occlusion 4 66     \nDouglas' lieaments DIE         \nNo 41 33 26.803 0.000 \nHemi 30 38     \nBilateral 7 51     \nRectal / sigmoid colon DIE         \nNo 67 77 12.250 0.000 \nYes 11 45     \nDIE of the serosal layer of the posterior uterine wall         \nNo 71 86 11.887 0.001 \nYes 7 36     \nAnterior vaginal fornix with tenderness         \nNo 77 120 0.041 0.839 \nYes 1 2     \nThe posterior vaginal fornix was tender         \nNo 61 100 0.429 0.512 \nYes 17 22     \nThe left appendage area was tender         \nNo 53 80 0.120 0.729 \nYes 25 42     \nThe right accessory area was tender         \nNo 60 82 2.179 0.140 \nYes 18 40     \nThe left sacral ligament was tender         \nNo 65 92 1.770 0.183 \nYes 13 30     \nThe right sacral ligament was tender         \nNo 74 97 9.059 0.003 \nYes 4 25     \nTable 1. Baseline characteristics of patients. \nBMI: body mass index; DIE: deep infiltrating endometriosis. \n\nUnivariate analysis of the diagnostic efficacy of \nimaging characteristics \nBased on univariate analysis, we analyzed a total \nof 49 variables,  including age, the presence of             \nadenomyosis, uterine size, ovarian EMT, the            \ndisappearance of bilateral ovarian malposition and \nmotion, the degree  of obliteration of the                  \nrectouterine pouch, DIE size and DIE distribution in \nthe fallopian  tube, kidney, renal ureters, urethra, \nbladder, bladder uterine peritoneal reflection ,                \nrectouterine pouch, rectovaginal septum, rectum/\nsigmoid colon, serosa of the  posterior uterine wall \nand posterior fornix of the vagina. The results  of \nthe univariate analysis showed that the r-ASRM \nstage of EMT lesions was correlated  with the       \npresence of adenomyosis, uterine enlargement, the \noccurrence of unilateral  or bilateral ovarian          \nendometrioma, the disappearance of unilateral or \nbilateral ovarian malposition and motion, the degree \nof the rectouterine pouch obliteration , the degree of \nuterosacral ligament involvement, involvement of the \nrectum/ sigmoid colon, involvement of the serosa \nof the posterior uterine wall, the presence  of          \ntenderness in the right uterosacral ligament, the \nmaximum diameters of chocolate  cysts in the left \nand right ovaries, maximum diameters of DIE              \nlesions located  in the left and right uterosacral               \nligaments, the maximum diameter and infiltration  \ndepth of intestinal DIE lesions, and the maximum \ndiameter and infiltration depth of EMT lesions in the \nserosa of the posterior uterine wall (all P < 0.05, see \ntable 2 for details).  \n994 Int. J. Radiat. Res., Vol. 22 No. 4, October 2024 \nFigure 1. A representative transvaginal ultrasonography image \nshowed the kissing ovaries sign with bilateral ovarian             \nendometriotic cysts. \nFigure 2. Representative transvaginal ultrasonography images \nshowed right uterosacral ligament deep infiltrating                     \nendometriosis (DIE). \nFigure 3. A representative transvaginal ultrasonography image \nshowed the Indian headdress sign of intestinal DIE. \nFigure 4. Representative transvaginal ultrasonography images \nof a DIE patient with the obliteration of the rectouterine \npouch. The lesion adheres to surrounding tissues, and the \npresented uterine sliding sign indicated the obliteration of the \nrectouterine pouch. \nFigure 6. Right \novarian                      \nendometriotic cyst \n(“white arrow” as \nshown) and              \nintestinal DIE \n(“black arrow” as \nshown) under the \nhysteroscope. \nFigure 5. Bilateral \nuterosacral            \nligament in a DIE \npatient under the \nhysteroscope. \n\nMultivariate analysis of the diagnostic value of \nimaging characteristics \nMultivariate binary logistic regression analysis \nwas conducted with  r- ASRM Stage IV as the            \ndependent variable. Our results showed that the size \nof the right ovarian endometrioma ,  the occurrence \nof unilateral or bilateral ovarian endometrioma, \nand the  degree of obliteration of the rectouterine \npouch were independent predictors for r -ASRM \nstage in EMT patients  (P < 0.05, table 3). The                 \nfollowing logistic regression equation was                    \nestablished: −7.901 + (0.031 × maximum diameter \n[long diameter, unit: mm] of right ovarian                \nendometrioma) + (2.437 × unilateral or bilateral \novarian endometrioma occurrence) + (2.888 ×              \ndegree of rectouterine pouch obliteration). Table  4 \nshowed the assigned values for regression analysis, \nwith an intercept of − 0.337. \nThe receiver operating characteristic (ROC) curve \nwas prepared using the predicted values. The predic-\ntive ability of this model was evaluated using the \narea under the  ROC curve ( figure 7 ) .  The logistic \nregression model established based on the imaging  \ncharacteristics of ultrasonography showed  a sensi-\ntivity of 82.0%, a specificity of 93.6%, an accuracy \nof 86.5%, and an area under the curve of 0.933 \n(standard error 0.016, P < 0 .05, 95% confidence in-\nterval: 0.901, 0.965). \nTable 3. Binary logistic regression results. \nSu et al. / A model using transvaginal ultrasonography to predict EMT  995 \n  I-III (n=78) IV (n=122) T/Z P \nAge 33.62±5.87 34.33±7.61 -0.744 0.458 \nLong diameter of left ovary 0.00 (0.00, 57.25) 36.00 (0.00, 70.00) -2.538 0.011 \nLong diameter of right ovary 0.00 (0.00, 56.25) 48.50 (21.75, 72.25) -4.698 0.000 \nDIE diameter of left uterosacral ligament 0.00 (0.00, 10.00) 7.50 (0.00, 13.00) -2.434 0.015 \nDIE diameter of right uterosacral ligament 0.00 (0.00, 0.00) 9.00 (0.00, 15.00) -5.461 0.000 \nIntestinal DIE length 0.00 (0.00, 0.00) 0.00 (0.00, 16.00) -3.514 0.000 \nGut DIE depth 0.00 (0.00, 0.00) 0.00 (0.00, 6.00) -3.536 0.000 \nThe DIE length of the uterine serosal layer 0.00 (0.00, 0.00) 0.00 (0.00, 10.00) -3.436 0.001 \nThe DIE depth of the uterine serosal layer 0.00 (0.00, 0.00) 0.00 (0.00, 4.00) -3.412 0.001 \nTable 2. Univariate analysis of the diagnostic efficacy of imaging characteristics. \nDIE: deep infiltrating endometriosis. \n  B standard \nerror Wald conspicuousness OR \nThe 95% confidence \nintervals of the OR \nlower limit superior limit \nLeft of left size (long diameter in mm) 0.022 0.012 3.742 0.053 1.023 1.000 1.046 \nRight sac size (long diameter in mm) 0.031 0.013 5.978 0.014 1.032 1.006 1.058 \nLeft uterosacral ligament DIE diameter (in mm) 0.028 0.059 0.219 0.640 1.028 0.916 1.154 \nRight uterosacral ligament DIE diameter (in mm) 0.065 0.065 0.987 0.320 1.067 0.939 1.212 \nDIE length (in mm) 0.052 0.067 0.615 0.433 1.054 0.925 1.201 \nDIE depth (in mm) 0.004 0.139 0.001 0.979 1.004 0.764 1.318 \nDIE of uterine serous layer (in mm) 0.169 0.126 1.801 0.180 1.184 0.925 1.515 \nDIE depth of uterine serosa layer (in mm) -0.296 0.286 1.071 0.301 0.744 0.425 1.303 \nUterine adenomyosis -.593 0.655 0.820 0.365 0.553 0.153 1.994 \nuterine size 1.055 0.825 1.635 0.201 2.871 0.570 14.456 \nOvarian lesions 2.437 0.770 10.025 0.002 11.440 2.531 51.714 \nOvarian dislocation movement -.385 0.466 0.682 0.409 0.680 0.273 1.697 \nRectal notch with or without occlusion 2.888 0.585 24.407 0.000 17.951 5.709 56.446 \nDouglas' lieaments DIE 0.420 0.739 0.322 0.570 1.521 0.357 6.476 \nA DIE of the rectosigmoid colon -0.976 1.410 0.480 0.489 0.377 0.024 5.968 \nDIE of the serosal layer of the posterior uterine wall 0.201 1.915 0.011 0.916 1.222 0.029 52.103 \nThe right sacral ligament was tender 0.347 0.927 0.140 0.708 1.415 0.230 8.709 \nOR: odds ratio; DIE: deep infiltrating endometriosis. \nTest the outcome variable AUC standard errora The asymptotic \nsignificanceb \nAsymptotic 95% confidence interval \nlower limit superior limit \nRight sac size (long diameter in mm) 0.694 0.039 0.000 0.618 0.770 \nOvarian lesions 0.774 0.033 0.000 0.710 0.838 \nRectal notch with or without occlusion 0.849 0.028 0.000 0.795 0.903 \nprediction model 0.933 0.016 0.000 0.901 0.965 \nTable 4. The area below the curve.  \nAUC: area under the curve. \n\n \n \n \n \n \n \n \n \n \n \n \n \n \n \nDISCUSSION \n \nEMT-induced chronic pelvic pain and infertility \nseriously affect the quality of life  of childbearing \nwomen globally. The most important role of r-ASRM \nclassification is to predict the postoperative            \ncapabilities of natural pregnancy in patients with \nEMT and provide  treatment options ( 1 5 ) .             \nLaparoscopic resection of EMT lesions is the most \ncommon treatment method for EMT. However, the \nprocedure is challenging, highly dependent  on the \nclinical experience and surgical skills of surgeons  \nand may trigger postoperative  complications/\ntrauma. Thus, physicians are suggested to consider \nthe preoperative imaging results to provide                  \nindividualized treatment plans. With the development \nof high -frequency endo-cavity probes, more imaging \ndetails are obtained with pelvic ultrasound                      \ninformation for the diagnosis of EMT. \nPrevious studies have indicated that the                \nadenomyosis is associated with endometriosis,             \nespecially the deep infiltrating lesions, and the             \nseverity of EMT was evaluated based on ASRM on \npreoperative transvaginal ultrasonography, and the \nresults of transvaginal ultrasonography are closely \ncorrelated with those of the laparoscopic                    \nexamination (16,17). Moreover, adenomyosis, right  \nendometrioma, right endometrioma ≥ 5 cm are     \nregarded as independent risk factors for EMT (18). A \nprevious work also revealed that the transvaginal \nultrasound sliding sign shows high sensitivity in the \nprediction of pouch of Douglas obliteration (19). In the \npresent study, we retrospectively analyzed the              \nclinical data and ultrasonography  imaging                      \ncharacteristics of EMT patients with confirmed                \nr-ASRM stages after laparoscopic surgery. Our results \nshowed that patients with adenomyosis and  uterine \nenlargement showed severe EMT and a high risk of \ninvolvement of bilateral  ovaries, bilateral                   \nuterosacral ligaments, the rectum/ sigmoid colon, \nand the serosa of the posterior uterine wall (all P < \n0.05), resulting in the disappearance of bilateral  \novarian malposition and motion, complete             \nobliteration of the rectouterine pouch, and                    \ntenderness in the right uterosacral ligament. T hree \nindependent predictors were  screened out after  \nunivariate analysis, including the maximum               \ndiameter of the right  ovarian endometrioma, the  \ndegree of obliteration of the rectouterine pouch, and \nunilateral or bilateral ovarian involvement (table 2) \nand the results are in line with the previous findings. \nThe classical theory of EMT pathogenesis is        \nretrograde menstruation, in which endometrial ,  \nepithelial and stromal cells in the blood flow back \ninto the pelvic cavity instead  of the vagina,                    \nstimulating the proliferation of connective tissue \nor smooth muscle tissue to form lesions ( 2 0 , 2 1 ). Also, \nanatomical and hormonal factors are considered  to \naffect left lateral predisposition to EMT. For             \nexample, the presence of a sigmoid  colon reduces \nthe blood flow to the left side of the pelvic cavity, \nwhich may delay the clearance of endometrial cells \nthrough tubal reflux during menstruation,                 \nsupporting the retrograde menstruation theory (22-24). \nIn th e present study, the maximum  diameter of the \nendometrioma in the right but not the left ovary had \npredictive power (table 3), which was consistent with \nthe findings reported by Ulukus et al . that the               \nincidence of ovarian endometrioma on the right side \nwas higher than that on the left side in patients with \nsevere EMT (25). In the r-ASRM classification, the         \ndegree of adhesions  on the ovaries greatly                 \ncontributes to the r-ASRM score. However, ovarian \nmalposition and motion had no independent                \npredictive value in our study. This may be related  to \nthe influence of subjective factors, such as the           \nimaging methods and the operator’s judgment. \nA multicenter study reveals that the transvaginal \nultrasound is of high accuracy in the prediction of \nASRM staging for EMT (26). A recent study by Yang et \nal. have established a preoperative prediction model \nfor the evaluation of risk factors related to severe \nEMT, with good diagnostic performance (AUC = \n0.846) (27).  In  the present study, we combined          \nmultiple imaging characteristics of ultrasonography  \nand clinical-related factors to establish a model for \npredicting severe EMT  (r-ASRM classification stage \nIV) via laparoscopic visualization through                     \nmultivariate logistic regression analysis. The           \nestablished model had a sensitivity of 82.0%, a \nspecificity of 93.6%, an accuracy of 86.5%, and an \narea under the curve of 0 .933 ( standard error \n0.016, P < 0.05, 95% confidence interval: 0.901, \n0.965) in predicting  severe EMT (table 4, figure 7) . \nThe predictive variables were all easy-to-obtain             \nfeatures that did not require invasive examinations. \n996 Int. J. Radiat. Res., Vol. 22 No. 4, October 2024 \nFigure 7. Receiver operating characteristic (ROC) curve              \nevaluates the predictive ability of the logistic regression model \nestablished in this study for severe endometriosis. The curve \nshows an area under the curve of 0.933, a standard error of \n0.016, P < 0.05 and a 95% confidence interval of (0.901, \n0.965). \n\nOur established model was simple, practical and \neasy to opera te, with good performance relative to \nthe previous models  and could be applied in a                  \nreal-time, fast, and repetitive manner in clinical             \npractice. \nIn the era of precision medicine, a single               \nindicator can no longer fulfill individualized                     \ntreatment needs. It is necessary to comprehensively \nanalyze the available information and screen out the \nrisk factors to improve diagnostic accuracy. The  \nprediction model established in this study is an             \neasy-to-use tool with clinical value. The risk of          \nsevere EMT in each patient was determined using \na simple equation,  facilitating individualized                \ntreatment and assisting physicians in accurately            \ndetermining each  patient's disease severity and               \nsurgical difficulties. \nNevertheless, this study has some limitations. \nElastography and contrast -enhanced                               \nultrasonography have rapidly developed in recent \nyears. Diagnostic ultrasound is no longer limited \nto simple grayscale and color Doppler imaging. \nMoreover, MRI has a certain value in the diagnosis \nand preoperative staging of  EMT (28) . These factors \nwere not included in this study. Follow-up research \nwith larger  samples should be conducted to further \nestablish a prediction model in combination  with \nmultimodal imaging indicators to verify th e                \naccuracy and practicability of our model. \n \n \nCONCLUSION \n \nTransvaginal ultrasonography is valuable for \nEMT staging.  The regression model established in \nthis study using ultrasonic imaging characteristics  \neffectively predicted the r- ASRM stage of EMT  \nlesions, providing a basis  for the diagnosis and \ntreatment of EMT, assisting surgeons in predicting \nthe degree  of surgical risk before surgery accurately \nand improving individualized surgical treatment. \n \nACKNOWLEDGMENT \nNot applicable. \n \nFunding: This article is one of the achievements of \nthe medical research project \"Application of Endo-\nscopic Ultrasound Elastic Imaging Technology Com-\nbined with Color Doppler in the Diagnosis of Deep \nLocalized Endometriosis of the Pelvic Cavity\" by the \nFoshan Municipal Health Bureau (project approval \nnumber 20210394). \nConflict of interests: All authors declared no conflicts \nof interest. \nData availability statements: The data that              \nsupport the findings of this study are available on \nrequest from the corresponding a uthor upon                 \nreasonable request. \nEthical consideration: Not applicable. \nAuthor contribution:  Q.S.: Conceptualization, data \nanalysis, original draft; H.L., J.G.: H.N.: data collection, \ndata analysis, reviewing and editing; Z.X., C.Z., J.C.: \ndata collection, data analysis, reviewing; P.W.:                  \nConceptualization, Supervision, reviewing and                \nediting. All authors have read and approved the final \nversion of the manuscript. \n  \n \nREFERENCES \n \n1. Shafrir AL, Farland LV, Shah DK , et al.  (2018) Risk for and conse-\nquences of endometriosis: A  critical epidemiologic review. Best \nPract Res Clin Obstet Gynaecol, 51: 1-15. \n2. Evans MB and Decherney AH (2017) Fertility and endometriosis. \nClin Obstet Gynecol, 60(3): 497-502. \n3. Bafort C, Beebeejaun Y, Tomassetti C , et al.  (2020) Laparoscopic \nsurgery for endometriosis. Cochrane Database Syst Rev , 10(10): \nCd011031. \n4. Johnson NP, Hummelshoj L, Adamson GD, et al. (2017) World endo-\nmetriosis society consensus on the classification of endometriosis. \nHum Reprod, 32(2): 315-324. \n5. Tomassetti C, Bafort C, Vanhie A , et al.  (2021) Estimation of the \nendometriosis fertility Index prior to operative laparoscopy. Hum \nReprod, 36(3): 636-646. \n6. Kido A, Himoto Y, Moribata Y , et al.  (2022) MRI in the diagnosis of \nendometriosis and related diseases. Korean Journal of Radiology , \n23(4): 426-445. \n7. Bazot M and Daraï E (2017) Diagnosis of deep endometriosis: Clini-\ncal examination, ultrasonography, magnetic resonance imaging \nand other techniques. Fertility and Sterility, 108(6): 886-894. \n8. Chapron C, Marcellin L, Borghese B, et al. (2019) Rethinking mecha-\nnisms, diagnosis and management of endometriosis. Nature Re-\nviews Endocrinology, 15(11): 666-682. \n9. Deslandes A, Parange N, Childs JT , et al.  (2020) Current status of \ntransvaginal ultrasound accuracy in the diagnosis of deep infil-\ntrating endometriosis before surgery: A systematic review of the \nliterature. J Ultrasound Med, 39(8): 1477-1490. \n10. Guerriero S and Ajossa S (2020) Transvaginal ultrasonography in \nsuperficial endometriosis. Aust N Z J Obstet Gynaecol, 60(3): E5. \n11. Guerriero S, Saba L, Pascual MA , et al.  (2018) Transvaginal ultra-\nsound vs magnetic resonance imaging for diagnosing deep infil-\ntrating endometriosis: Systematic review and meta-analysis. Ultra-\nsound in Obstetrics & Gynecology: The Official Journal of the Inter-\nnational Society of Ultrasound in Obstetrics and Gynecology, 51(5): \n586-595. \n12. Aas-Eng MK, Young VS, Dormagen JB , et al. (2023) Lesion-to-anal-\nverge distance in rectosigmoid endometriosis on transvaginal \nsonography vs magnetic resonance imaging: Prospective study. \nUltrasound in Obstetrics and Gynecology: The Official Journal of \nthe International Society of Ultrasound in Obstetrics and Gynecolo-\ngy, 61(2): 243-250. \n13. Lee MS, Kim MD, Lee M , et al.  (2012) Contrast -enhanced MR \nangiography of uterine arteries for the prediction of ovarian artery \nembolization in 349 patients. Journal of Vascular and Intervention-\nal Radiology: JVIR, 23(9): 1174-1179. \n14. Indrielle-Kelly T, Frühauf F, Fanta M , et al.  (2020) Application of \ninternational deep endometriosis analysis (IDEA) group consensus \nin preoperative ultrasound and magnetic resonance imaging of \ndeep pelvic endometriosis. Ultrasound Obstet Gynecol, 56(1): 115-\n116. \n15. Keckstein J, Oppelt P, Hudelist G (2022) Classification and clinical \nstaging of endometriosis. In: Oral E, editor. Endometriosis and \nadenomyosis: Global perspectives across the lifespan. Cham: \nSpringer International Publishing, 93-108. \n16. Chapron C, Tosti C, Marcellin L, et al. (2017) Relationship between \nthe magnetic resonance imaging appearance of adenomyosis and \nendometriosis phenotypes. Human Reproduction , 32(7): 1393-\n1401. \n17. Holland TK, Cutner A, Saridogan E , et al.  (2013) Ultrasound map-\nping of pelvic endometriosis: Does the location and number of \nlesions affect the diagnostic accuracy? A multicentre diagnostic \naccuracy study. BMC Women's Health, 13: 43. \n18. Mabrouk M, Raimondo D, Mastronardi M, et al. (2020) Endometri-\nosis of the appendix: When to predict and how to manage-A multi-\nSu et al. / A model using transvaginal ultrasonography to predict EMT  997 \n\nvariate analysis of 1935 endometriosis cases. Journal of Minimally \nInvasive Gynecology, 27(1): 100-106. \n19. Arion K, Aksoy T, Allaire C , et al.  (2019) Prediction of pouch of \ndouglas obliteration: Point-of-care ultrasound versus pelvic exami-\nnation. Journal of Minimally Invasive Gynecology, 26(5): 928-934. \n20. Borghese B, Santulli P, Marcellin L , et al.  (2018) [Definition, de-\nscription, clinicopathological features, pathogenesis and natural \nhistory of endometriosis: CNGOF -HAS Endometriosis Guidelines]. \nGynecol Obstet Fertil Senol, 46(3): 156-167. \n21. Koninckx PR, Fernandes R, Ussia A , et al.  (2021) Pathogenesis \nbased diagnosis and treatment of endometriosis. Front Endocrinol \n(Lausanne), 12: 745548. \n22. Vercellini P, Aimi G, De Giorgi O , et al.  (1998) Is cystic ovarian \nendometriosis an asymmetric disease? Br J Obstet Gynaecol , 105\n(9): 1018-1021. \n23. Al-Fozan H and Tulandi T (2003) Left lateral predisposition of en-\ndometriosis and endometrioma. Obstet Gynecol, 101(1): 164-166. \n24. Sznurkowski JJ and Emerich J (2008) Endometriomas are more \nfrequent on the left side. Acta Obstet Gynecol Scand , 87(1): 104-\n106. \n25. Ulukus M, Yeniel A, Ergenoglu AM , et al. (2012) Right endometrio-\nma is related with more extensive obliteration of the Douglas \npouch. Arch Gynecol Obstet, 285(5): 1483-1486. \n26. Leonardi M, Espada M, Choi S , et al.  (2020) Transvaginal ultra-\nsound can accurately predict the american society of reproductive \nmedicine stage of endometriosis assigned at laparoscopy. Journal \nof Minimally Invasive Gynecology, 27(7): 1581-7. e1. \n27. Yang Y, Li J, Chen H , et al. (2022) Assessment of risk factors associ-\nated with severe endometriosis and establishment of preoperative \nprediction model. Diagnostics (Basel, Switzerland), 12(10). \n28. Bazot M, Kermarrec E, Bendifallah S , et al. (2021) MRI of intestinal \nendometriosis. Best Pract Res Clin Obstet Gynaecol, 71: 51-63.  \n998 Int. J. Radiat. Res., Vol. 22 No. 4, October 2024","source_license":"CC0","license_restricted":false}