E. Epstein

Combination of Proactive Molecular Risk Classifier for Endometrial cancer (ProMisE) with sonographic and demographic characteristics in preoperative prediction of recurrence or progression of endometrial cancer

ABSTRACTObjectiveTo evaluate the ability of demographic and sonographic variables and the Proactive Molecular Risk Classifier for Endometrial cancer (ProMisE) classification to predict preoperatively tumor recurrence or progression in women with endometrial cancer.MethodsThe study included 339 women with histologically confirmed endometrial cancer who underwent expert transvaginal ultrasound in a single center before surgery as part of the prospective International Endometrial Tumor Analysis 4 study or who were evaluated using the same protocol. The tumors were classified according to histotype, FIGO (International Federation of Gynecology and Obstetrics) grade and FIGO stage. In addition, molecular analysis was performed for classification into the four ProMisE subtypes: polymerase‐ϵ exonuclease domain mutations (POLE EDM), mismatch repair proteins deficiency (MMR‐D), protein 53 wild type (p53 wt) and protein 53 abnormal (p53 abn). Demographic and preoperative sonographic characteristics, tumor recurrence or progression and survival were compared between the ProMisE subgroups. Cox regression analysis was used to identify prognostic factors associated with recurrence or progression, using univariable models to study crude associations and multivariable models to study adjusted associations. Logistic regression and receiver‐operating‐characteristics (ROC)‐curve analysis were used to assess the predictive ability of the preoperative prognostic factors regarding recurrence or progression of cancer within 3 years after surgery, and to compare their predictive ability to that of the European Society for Medical Oncology (ESMO) preoperative (based on depth of myometrial invasion, histotype and grade) and postoperative (based on histotype, grade, surgical stage and lymphovascular space invasion) risk classifications. In a separate subanalysis, cases were stratified according to ProMisE p53 abn status (present vs absent) and sonographic tumor size (anteroposterior (AP) diameter < 2 cm vs ≥ 2 cm).ResultsMedian follow‐up time from surgery was 58 months (interquartile range, 48–71 months; range, 0–102 months). Recurrence or progression of cancer occurred in 51/339 (15%) women, comprising 14% of those with MMR‐D, 8% of those with POLE EDM, 9% of those with p53 wt and 45% of those with p53 abn ProMisE subtype. On multivariable analysis, age, waist circumference, ProMisE subtype and tumor extension and AP diameter on ultrasound were associated with tumor recurrence or progression. A multivariable model comprising ProMisE subtype, age, waist circumference and sonographic tumor extension and size (area under the ROC curve (AUC), 0.89 (95% CI, 0.85–0.93)) had comparable ability to predict tumor recurrence/progression to that of a multivariable model comprising histotype, grade, age, waist circumference and sonographic tumor extension and size (AUC, 0.88 (95% CI, 0.83–0.92)), and better predictive ability than both the preoperative (AUC, 0.74 (95% CI, 0.67–0.82); P < 0.01) and postoperative (AUC, 0.79 (95% CI, 0.72–0.86); P < 0.01) ESMO risk classifications. Women with a combination of non‐p53 abn subtype and tumor size < 2 cm (164/339 (48%)) had a very low risk (1.8%) of tumor recurrence or progression.ConclusionsThe combination of demographic characteristics, sonographic findings and ProMisE subtype had better preoperative predictive ability for tumor recurrence or progression than did the ESMO classification, supporting their use in the preoperative risk stratification of women with endometrial cancer. The combination of p53 status with ultrasound tumor size has the potential to identify preoperatively a large group of women with a very low risk of recurrence or progression. © 2020 The Authors. Ultrasound in Obstetrics & Gynecology published by John Wiley & Sons Ltd on behalf of International Society of Ultrasound in Obstetrics and Gynecology. ‐ Legal Statement: This is an open access article under the terms of the Creative Commons Attribution License, which permits use, distribution and reproduction in any medium, provided the original work is properly cited.

Ultrasound‐based risk model for preoperative prediction of lymph‐node metastases in women with endometrial cancer: model‐development study

ABSTRACTObjectiveTo develop a preoperative risk model, using endometrial biopsy results and clinical and ultrasound variables, to predict the individual risk of lymph‐node metastases in women with endometrial cancer.MethodsA mixed‐effects logistic regression model for prediction of lymph‐node metastases was developed in 1501 prospectively included women with endometrial cancer undergoing transvaginal ultrasound examination before surgery, from 16 European centers. Missing data, including missing lymph‐node status, were imputed. Discrimination, calibration and clinical utility of the model were evaluated using leave‐center‐out cross validation. The predictive performance of the model was compared with that of risk classification from endometrial biopsy alone (high‐risk defined as endometrioid cancer Grade 3/non‐endometrioid cancer) or combined endometrial biopsy and ultrasound (high‐risk defined as endometrioid cancer Grade 3/non‐endometrioid cancer/deep myometrial invasion/cervical stromal invasion/extrauterine spread).ResultsLymphadenectomy was performed in 691 women, of whom 127 had lymph‐node metastases. The model for prediction of lymph‐node metastases included the predictors age, duration of abnormal bleeding, endometrial biopsy result, tumor extension and tumor size according to ultrasound and undefined tumor with an unmeasurable endometrium. The model's area under the curve was 0.73 (95% CI, 0.68–0.78), the calibration slope was 1.06 (95% CI, 0.79–1.34) and the calibration intercept was 0.06 (95% CI, –0.15 to 0.27). Using a risk threshold for lymph‐node metastases of 5% compared with 20%, the model had, respectively, a sensitivity of 98% vs 48% and specificity of 11% vs 80%. The model had higher sensitivity and specificity than did classification as high‐risk, according to endometrial biopsy alone (50% vs 35% and 80% vs 77%, respectively) or combined endometrial biopsy and ultrasound (80% vs 75% and 53% vs 52%, respectively). The model's clinical utility was higher than that of endometrial biopsy alone or combined endometrial biopsy and ultrasound at any given risk threshold.ConclusionsBased on endometrial biopsy results and clinical and ultrasound characteristics, the individual risk of lymph‐node metastases in women with endometrial cancer can be estimated reliably before surgery. The model is superior to risk classification by endometrial biopsy alone or in combination with ultrasound. Copyright © 2019 ISUOG. Published by John Wiley & Sons Ltd.

Dynamic contrast‐enhanced ultrasound improves diagnostic performance in endometrial cancer staging

ABSTRACTObjectivesTo compare the sensitivity and specificity of conventional two‐dimensional transvaginal ultrasound/power Doppler (2D‐TVU/PD) alone and 2D‐TVU/PD combined with dynamic contrast‐enhanced ultrasound (DCE‐US) in diagnosing deep myometrial invasion (MI) and cervical stromal involvement (CSI) in women with endometrial cancer (EC), and to assess the association of DCE‐US semiquantitative and qualitative variables with International Federation of Gynecology and Obstetrics (FIGO) Stage ≥ IB and ‘high‐risk’ cancer.MethodsThis was a prospective study of 101 consecutive women with biopsy‐confirmed EC, undergoing expert ultrasound examination at Karolinska University Hospital, a tertiary referral center. All consenting women underwent DCE‐US using a 1.5–2.5‐mL intravenous bolus of SonoVue contrast agent, as well as conventional 2D‐TVU/PD examination. DCE‐US videoclips were analyzed with regard to filling (global or focal), wash‐in (prior, simultaneous or after) and wash‐out (global or focal) patterns of the contrast agent in the tumor compared with the surrounding tissue, as well as semiquantitative DCE‐US parameters (wash‐in slope, time‐to‐peak, peak intensity and area under the time–intensity curve (TIC)) obtained from a TIC. The study cohort was compared with a control cohort of women with EC examined at our center according to the International Endometrial Tumor Analysis protocol using 2D‐TVU/PD only, matched at a ratio of 3:1 for FIGO stage and grade. The sensitivity and specificity of 2D‐TVU/PD alone in the control cohort and in combination with DCE‐US in the study cohort in the diagnosis of deep MI, CSI and high‐risk cancer (defined as FIGO Stage ≥ IB and/or Grade 3 endometrioid and/or non‐endometrioid histology) were compared, using pathological evaluation after hysterectomy as the ‘gold standard’.ResultsAfter exclusions, 93 women were included in the study cohort and were matched to 279 women in the control cohort. The prevalence of FIGO Stage IA, Grade 1–2 EC was 52% in both cohorts. The sensitivity of 2D‐TVU/PD with DCE‐US in the study cohort was higher than that of 2D‐TVU/PD alone in the control cohort in diagnosing both deep MI (0.74 vs 0.62; P = 0.036) and CSI (0.75 vs 0.51; P < 0.001), whereas the specificity was not significantly different (0.87 vs 0.85 and 0.96 vs 0.95, respectively). Compared with 2D‐TVU/PD alone, the specificity of 2D‐TVU/PD with DCE‐US was higher in detecting high‐risk cancer (0.94 vs 0.85; P = 0.024) but the sensitivity did not differ (0.73 vs 0.71). High‐risk cancer and FIGO Stage ≥ IB were characterized by a ‘focal’ filling pattern, with a ‘prior’ wash‐in pattern and a ‘focal’ wash‐out pattern on subjective assessment of DCE‐US videoclips. All semiquantitative DCE‐US parameters were significantly predictive of FIGO Stage ≥ IB but not of high‐risk cancer, despite a clear trend.ConclusionsCompared with 2D‐TVU/PD alone, combining 2D‐TVU/PD with DCE‐US can significantly improve the detection of deep MI and CSI in women with EC, without increasing the false‐positive rate. It can also improve the correct classification of high‐risk disease, mainly by increasing specificity, thereby possibly reducing the number of unnecessarily extensive surgeries by almost 10%. Semiquantitative DCE‐US parameters, as well as a ‘focal’ filling pattern, endometrial wash‐in prior to the myometrium and a ‘focal’ wash‐out pattern, are all associated with more advanced disease. Copyright © 2019 ISUOG. Published by John Wiley & Sons Ltd.

Validation of ultrasound strategies to assess tumor extension and to predict high‐risk endometrial cancer in women from the prospective IETA (International Endometrial Tumor Analysis)‐4 cohort

ABSTRACTObjectivesTo compare the performance of ultrasound measurements and subjective ultrasound assessment (SA) in detecting deep myometrial invasion (MI) and cervical stromal invasion (CSI) in women with endometrial cancer, overall and according to whether they had low‐ or high‐grade disease separately, and to validate published measurement cut‐offs and prediction models to identify MI, CSI and high‐risk disease (Grade‐3 endometrioid or non‐endometrioid cancer and/or deep MI and/or CSI).MethodsThe study comprised 1538 patients with endometrial cancer from the International Endometrial Tumor Analysis (IETA)‐4 prospective multicenter study, who underwent standardized expert transvaginal ultrasound examination. SA and ultrasound measurements were used to predict deep MI and CSI. We assessed the diagnostic accuracy of the tumor/uterine anteroposterior (AP) diameter ratio for detecting deep MI and that of the distance from the lower margin of the tumor to the outer cervical os (Dist‐OCO) for detecting CSI. We also validated two two‐step strategies for the prediction of high‐risk cancer; in the first step, biopsy‐confirmed Grade‐3 endometrioid or mucinous or non‐endometrioid cancers were classified as high‐risk cancer, while the second step encompassed the application of a mathematical model to classify the remaining tumors. The ‘subjective prediction model’ included biopsy grade (Grade 1 vs Grade 2) and subjective assessment of deep MI or CSI (presence or absence) as variables, while the ‘objective prediction model’ included biopsy grade (Grade 1 vs Grade 2) and minimal tumor‐free margin. The predictive performance of the two two‐step strategies was compared with that of simply classifying patients as high risk if either deep MI or CSI was suspected based on SA or if biopsy showed Grade‐3 endometrioid or mucinous or non‐endometrioid histotype (i.e. combining SA with biopsy grade). Histological assessment from hysterectomy was considered the reference standard.ResultsIn 1275 patients with measurable lesions, the sensitivity and specificity of SA for detecting deep MI was 70% and 80%, respectively, in patients with a Grade‐1 or ‐2 endometrioid or mucinous tumor vs 76% and 64% in patients with a Grade‐3 endometrioid or mucinous or a non‐endometrioid tumor. The corresponding values for the detection of CSI were 51% and 94% vs 50% and 91%. Tumor AP diameter and tumor/uterine AP diameter ratio showed the best performance for predicting deep MI (area under the receiver–operating characteristics curve (AUC) of 0.76 and 0.77, respectively), and Dist‐OCO had the best performance for predicting CSI (AUC, 0.72). The proportion of patients classified correctly as having high‐risk cancer was 80% when simply combining SA with biopsy grade vs 80% and 74% when using the subjective and objective two‐step strategies, respectively. The subjective and objective models had an AUC of 0.76 and 0.75, respectively, when applied to Grade‐1 and ‐2 endometrioid tumors.ConclusionsIn the hands of experienced ultrasound examiners, SA was superior to ultrasound measurements for the prediction of deep MI and CSI of endometrial cancer, especially in patients with a Grade‐1 or ‐2 tumor. The mathematical models for the prediction of high‐risk cancer performed as expected. The best strategies for predicting high‐risk endometrial cancer were combining SA with biopsy grade and the subjective two‐step strategy, both having an accuracy of 80%. Copyright © 2019 ISUOG. Published by John Wiley & Sons Ltd.

Ultrasound image analysis using deep neural networks for discriminating between benign and malignant ovarian tumors: comparison with expert subjective assessment

AbstractObjectivesTo develop and test the performance of computerized ultrasound image analysis using deep neural networks (DNNs) in discriminating between benign and malignant ovarian tumors and to compare its diagnostic accuracy with that of subjective assessment (SA) by an ultrasound expert.MethodsWe included 3077 (grayscale, n = 1927; power Doppler, n = 1150) ultrasound images from 758 women with ovarian tumors, who were classified prospectively by expert ultrasound examiners according to IOTA (International Ovarian Tumor Analysis) terms and definitions. Histological outcome from surgery (n = 634) or long‐term (≥ 3 years) follow‐up (n = 124) served as the gold standard. The dataset was split into a training set (n = 508; 314 benign and 194 malignant), a validation set (n = 100; 60 benign and 40 malignant) and a test set (n = 150; 75 benign and 75 malignant). We used transfer learning on three pre‐trained DNNs: VGG16, ResNet50 and MobileNet. Each model was trained, and the outputs calibrated, using temperature scaling. An ensemble of the three models was then used to estimate the probability of malignancy based on all images from a given case. The DNN ensemble classified the tumors as benign or malignant (Ovry‐Dx1 model); or as benign, inconclusive or malignant (Ovry‐Dx2 model). The diagnostic performance of the DNN models, in terms of sensitivity and specificity, was compared to that of SA for classifying ovarian tumors in the test set.ResultsAt a sensitivity of 96.0%, Ovry‐Dx1 had a specificity similar to that of SA (86.7% vs 88.0%; P = 1.0). Ovry‐Dx2 had a sensitivity of 97.1% and a specificity of 93.7%, when designating 12.7% of the lesions as inconclusive. By complimenting Ovry‐Dx2 with SA in inconclusive cases, the overall sensitivity (96.0%) and specificity (89.3%) were not significantly different from using SA in all cases (P = 1.0).ConclusionUltrasound image analysis using DNNs can predict ovarian malignancy with a diagnostic accuracy comparable to that of human expert examiners, indicating that these models may have a role in the triage of women with an ovarian tumor. © 2020 The Authors. Ultrasound in Obstetrics & Gynecology published by John Wiley & Sons Ltd on behalf of International Society of Ultrasound in Obstetrics and Gynecology.

Imaging in gynecological disease (22): clinical and ultrasound characteristics of ovarian embryonal carcinomas, non‐gestational choriocarcinomas and malignant mixed germ cell tumors

ABSTRACTObjectiveTo describe the clinical and ultrasound characteristics of three types of rare malignant ovarian germ cell tumor: embryonal carcinoma, non‐gestational choriocarcinoma and malignant mixed germ cell tumor.MethodsThis was a retrospective multicenter study. From the International Ovarian Tumor Analysis (IOTA) database, we identified patients with a histological diagnosis of ovarian embryonal carcinoma, non‐gestational choriocarcinoma or malignant mixed germ cell tumor, who had undergone preoperative ultrasound examination by an experienced ultrasound examiner between 2000 and 2020. Additional patients with the same histology were identified from the databases of the departments of gynecological oncology in the participating centers. All tumors were described using IOTA terminology. Three examiners reviewed all available ultrasound images and described them using pattern recognition.ResultsOne patient with embryonal carcinoma, five patients with non‐gestational ovarian choriocarcinoma and seven patients with ovarian malignant mixed germ cell tumor (six primary tumors and one recurrence) were identified. Seven patients were included in the IOTA studies and six patients were examined outside of the IOTA studies. The median age at diagnosis was 26 (range, 14–77) years. Beta‐human chorionic gonadotropin levels were highest in non‐gestational choriocarcinomas and alpha‐fetoprotein levels were highest in malignant mixed germ cell tumors. Most tumors were International Federation of Gynecology and Obstetrics (FIGO) Stage I (9/12 (75.0%)). All tumors were unilateral, and the median largest diameter was 129 (range, 38–216) mm. Of the tumors, 11/13 (84.6%) were solid and 2/13 (15.4%) were multilocular‐solid; 9/13 (69.2%) manifested abundant vascularization on color Doppler examination. Using pattern recognition, the typical ultrasound appearance was a large solid tumor with inhomogeneous echogenicity of the solid tissue and often dispersed cysts which, in most cases, were small and irregular. Some tumors had smooth contours while others had irregular contours.ConclusionsA unilateral, large solid tumor with inhomogeneous echogenicity of the solid tissue and with dispersed small cystic areas in a young woman should raise the suspicion of a rare malignant germ cell tumor. This suspicion can guide the clinician to test tumor markers specific for malignant germ cell tumors. © 2020 International Society of Ultrasound in Obstetrics and Gynecology

Vessel morphology depicted by three‐dimensional power Doppler ultrasound as second‐stage test in adnexal tumors that are difficult to classify: prospective diagnostic accuracy study

ABSTRACTObjectivesTo assess whether vessel morphology depicted by three‐dimensional (3D) power Doppler ultrasound improves discrimination between benignity and malignancy if used as a second‐stage test in adnexal masses that are difficult to classify.MethodsThis was a prospective observational international multicenter diagnostic accuracy study. Consecutive patients with an adnexal mass underwent standardized transvaginal two‐dimensional (2D) grayscale and color or power Doppler and 3D power Doppler ultrasound examination by an experienced examiner, and those with a ‘difficult’ tumor were included in the current analysis. A difficult tumor was defined as one in which the International Ovarian Tumor Analysis (IOTA) logistic regression model‐1 (LR‐1) yielded an ambiguous result (risk of malignancy, 8.3% to 25.5%), or as one in which the ultrasound examiner was uncertain regarding classification as benign or malignant when using subjective assessment. Even when the ultrasound examiner was uncertain, he/she was obliged to classify the tumor as most probably benign or most probably malignant. For each difficult tumor, one researcher created a 360° rotating 3D power Doppler image of the vessel tree in the whole tumor and another of the vessel tree in a 5‐cm3 spherical volume selected from the most vascularized part of the tumor. Two other researchers, blinded to the patient's history, 2D ultrasound findings and histological diagnosis, independently described the vessel tree using predetermined vessel features. Their agreed classification was used. The reference standard was the histological diagnosis of the mass. The sensitivity of each test for discriminating between benign and malignant difficult tumors was plotted against 1 – specificity on a receiver‐operating‐characteristics diagram, and the test with the point furthest from the reference line was considered to have the best diagnostic ability.ResultsOf 2403 women with an adnexal mass, 376 (16%) had a difficult mass. Ultrasound volumes were available for 138 of these cases. In 79/138 masses, the ultrasound examiner was uncertain about the diagnosis based on subjective assessment, in 87/138, IOTA LR‐1 yielded an ambiguous result and, in 28/138, both methods gave an uncertain result. Of the masses, 38/138 (28%) were malignant. Among tumors that were difficult to classify as benign or malignant by subjective assessment, the vessel feature ‘densely packed vessels’ had the best discriminative ability (sensitivity 67% (18/27), specificity 83% (43/52)) and was slightly superior to subjective assessment (sensitivity 74% (20/27), specificity 60% (31/52)). In tumors in which IOTA LR‐1 yielded an ambiguous result, subjective assessment (sensitivity 82% (14/17), specificity 79% (55/70)) was superior to the best vascular feature, i.e. changes in the diameter of vessels in the whole tumor volume (sensitivity 71% (12/17), specificity 69% (48/70)).ConclusionVessel morphology depicted by 3D power Doppler ultrasound may slightly improve discrimination between benign and malignant adnexal tumors that are difficult to classify by subjective ultrasound assessment. For tumors in which the IOTA LR‐1 model yields an ambiguous result, subjective assessment is superior to vessel morphology as a second‐stage test. © 2020 The Authors. Ultrasound in Obstetrics & Gynecology published by John Wiley & Sons Ltd on behalf of International Society of Ultrasound in Obstetrics and Gynecology.

Imaging in gynecological disease (17): ultrasound features of malignant ovarian yolk sac tumors (endodermal sinus tumors)

ABSTRACTObjectiveTo describe the clinical and sonographic characteristics of malignant ovarian yolk sac tumors (YSTs).MethodsIn this retrospective multicenter study, we included 21 patients with a histological diagnosis of ovarian YST and available transvaginal ultrasound images and/or videoclips and/or a detailed ultrasound report. Ten patients identified from the International Ovarian Tumor Analysis (IOTA) studies had undergone a standardized preoperative ultrasound examination, by an experienced ultrasound examiner, between 1999 and 2016. A further 11 patients were identified through medical files, for whom ultrasound images were retrieved from local image workstations and picture archiving and communication systems. All tumors were described using IOTA terminology. The collected ultrasound images and videoclips were used by two observers for additional characterization of the tumors.ResultsAll cases were pure YSTs, except for one that was a mixed tumor (80% YST and 20% embryonal carcinoma). Median age at diagnosis was 25 (interquartile range (IQR), 19.5–30.5) years. Seventy‐six percent (16/21) of women had an International Federation of Gynecology and Obstetrics (FIGO) Stage I–II tumor at diagnosis. Fifty‐eight percent (11/19) of women felt pain during the ultrasound examination and one presented with ovarian torsion. Median serum α‐fetoprotein (S‐AFP) level was 4755 (IQR, 1071–25 303) µg/L and median serum CA 125 level was 126 (IQR, 35–227) kU/L. On ultrasound assessment, 95% (20/21) of tumors were unilateral. The median maximum tumor diameter was 157 (IQR, 107–181) mm and the largest solid component was 110 (IQR, 66–159) mm. Tumors were classified as either multilocular‐solid (10/21; 48%) or solid (11/21; 52%). Papillary projections were found in 10% (2/21) of cases. Most (20/21; 95%) tumors were well vascularized (color score, 3–4) and none had acoustic shadowing. Malignancy was suspected in all cases, except in the patient with ovarian torsion, who presented a tumor with a color score of 1, which was classified as probably benign. Image and videoclip quality was considered as adequate in 18/21 cases. On review of the images and videoclips, we found that all tumors contained both solid components and cystic spaces, and that 89% (16/18) had irregular, still fine‐textured and slightly hyperechoic solid tissue, giving them a characteristic appearance.ConclusionMalignant ovarian YSTs are often detected at an early stage, in young women usually in the second or third decade of life, presenting with pain and markedly elevated S‐AFP. On ultrasound, malignant ovarian YSTs are mostly unilateral, large and multilocular‐solid or solid, with fine‐textured slightly hyperechoic solid tissue and rich vascularization. © 2020 The Authors. Ultrasound in Obstetrics & Gynecology published by John Wiley & Sons Ltd on behalf of the International Society of Ultrasound in Obstetrics and Gynecology..

Imaging in gynecological disease (24): clinical and ultrasound characteristics of ovarian mature cystic teratomas

ABSTRACTObjectiveTo describe the clinical and ultrasound features of ovarian mature cystic teratomas (MCTs).MethodsThis was a retrospective study. From the International Ovarian Tumor Analysis (IOTA) database, we identified patients with a histologically confirmed diagnosis of MCT who had undergone transvaginal ultrasound examination between 1999 and 2016 (IOTA phases 1, 2, 3 and 5) in one of five centers. Ultrasound was performed by an experienced examiner who used the standardized IOTA examination technique and terminology. In addition to extracting data from the IOTA database, available two‐dimensional grayscale and color or power Doppler images were reviewed retrospectively to identify typical ultrasound features of MCT described previously and detect possible new features using pattern recognition. All images were reviewed by two independent examiners and further discussed with two ultrasound experts to reach consensus.ResultsIncluded in the study were 454 patients with histologically confirmed MCT. Median age was 33 (range, 8–90)  years and 66 (14.5%) patients were postmenopausal. Most MCTs were described by the original ultrasound examiner as unilocular (262/454 (57.7%)) or multilocular (70/454 (15.4%)) cysts with mixed echogenicity of cystic fluid (368/454 (81.1%)), acoustic shadowing (328/454 (72.2%)) and no or little vascularization on color Doppler (color score 1, 240/454 (52.9%); color score 2, 123/454 (27.1%)). The median largest lesion diameter was 66 (range, 15–310)  mm. A correct preoperative diagnosis of MCT was suggested by the original ultrasound examiner in 372/454 (81.9%) cases. On retrospective review of ultrasound images of 334 MCTs that had quality sufficient for assessment, ‘dots and/or lines’ and/or ‘echogenic white ball’ (typical features according to the literature) were present in 271/334 (81.1%) masses. We identified four new ultrasound features characteristic of MCT: ‘cotton wool tufts’, ‘mushroom cap sign’, ‘completely hyperechogenic lesion’ and ‘starry sky sign’. At least one classical or novel ultrasound feature was present in 315/334 (94.3%) MCTs. Twenty‐nine (8.7%) MCTs manifested vascularized solid tissue, of which seven exhibited no typical features.ConclusionWe provide a comprehensive overview of conventional and newly described ultrasound features of MCTs. Only a small proportion of MCTs did not manifest any of the typical features. © 2022 International Society of Ultrasound in Obstetrics and Gynecology.

Imaging in gynecological disease (23): clinical and ultrasound characteristics of ovarian carcinosarcoma

ABSTRACTObjectiveTo describe the clinical and ultrasound characteristics of ovarian carcinosarcoma.MethodsThis was a retrospective multicenter study. Patients with a histological diagnosis of ovarian carcinosarcoma, who had undergone preoperative ultrasound examination between 2010 and 2019, were identified from the International Ovarian Tumor Analysis (IOTA) database. Additional patients who were examined outside of the IOTA study were identified from the databases of the participating centers. The masses were described using the terms and definitions of the IOTA group. Additionally, two experienced ultrasound examiners reviewed all available images to identify typical ultrasound features using pattern recognition.ResultsNinety‐one patients with ovarian carcinosarcoma who had undergone ultrasound examination were identified, of whom 24 were examined within the IOTA studies and 67 were examined outside of the IOTA studies. Median age at diagnosis was 66 (range, 33–91) years and 84/91 (92.3%) patients were postmenopausal. Most patients (67/91, 73.6%) were symptomatic, with the most common complaint being pain (51/91, 56.0%). Most tumors (67/91, 73.6%) were International Federation of Gynecology and Obstetrics (FIGO) Stage III or IV. Bilateral lesions were observed on ultrasound in 46/91 (50.5%) patients. Ascites was present in 38/91 (41.8%) patients. The median largest tumor diameter was 100 (range, 18–260) mm. All ovarian carcinosarcomas contained solid components, and most were described as solid (66/91, 72.5%) or multilocular‐solid (22/91, 24.2%). The median diameter of the largest solid component was 77.5 (range, 11–238) mm. Moderate or rich vascularization was found in 78/91 (85.7%) cases. Retrospective analysis of ultrasound images and videoclips using pattern recognition in 73 cases revealed that all tumors had irregular margins and inhomogeneous echogenicity of the solid components. Forty‐seven of 73 (64.4%) masses appeared as a solid tumor with cystic areas. Cooked appearance of the solid tissue was identified in 28/73 (38.4%) tumors. No pathognomonic ultrasound sign of ovarian carcinosarcoma was found.ConclusionsOvarian carcinosarcomas are usually diagnosed in postmenopausal women and at an advanced stage. The most common ultrasound appearance is a large solid tumor with irregular margins, inhomogeneous echogenicity of the solid tissue and cystic areas. The second most common pattern is a large multilocular‐solid mass with inhomogeneous echogenicity of the solid tissue. © 2021 International Society of Ultrasound in Obstetrics and Gynecology.

Terms, definitions and measurements to describe sonographic features of lymph nodes: consensus opinion from the Vulvar International Tumor Analysis (VITA) group

ABSTRACTIn centers with access to high‐end ultrasound machines and expert sonologists, ultrasound is used to detect metastases in regional lymph nodes from melanoma, breast cancer and vulvar cancer. There is, as yet, no international consensus on ultrasound assessment of lymph nodes in any disease or medical condition. The lack of standardized ultrasound nomenclature to describe lymph nodes makes it difficult to compare results from different ultrasound studies and to find reliable ultrasound features for distinguishing non‐infiltrated lymph nodes from lymph nodes infiltrated by cancer or lymphoma cells. The Vulvar International Tumor Analysis (VITA) collaborative group consists of gynecologists, gynecologic oncologists and radiologists with expertise in gynecologic cancer, particularly in the ultrasound staging and treatment of vulvar cancer. The work herein is a consensus opinion on terms, definitions and measurements which may be used to describe inguinal lymph nodes on grayscale and color/power Doppler ultrasound. The proposed nomenclature need not be limited to the description of inguinal lymph nodes as part of vulvar cancer staging; it can be used to describe peripheral lymph nodes in general, as well as non‐peripheral (i.e. parietal or visceral) lymph nodes if these can be visualized clearly. The association between the ultrasound features described here and histopathological diagnosis has not yet been established. VITA terms and definitions lay the foundations for prospective studies aiming to identify ultrasound features typical of metastases and other pathology in lymph nodes and studies to elucidate the role of ultrasound in staging of vulvar and other malignancies. © 2021 International Society of Ultrasound in Obstetrics and Gynecology.

Interobserver agreement of transvaginal ultrasound and magnetic resonance imaging in local staging of cervical cancer

ABSTRACTObjectiveTo evaluate interobserver agreement for the assessment of local tumor extension in women with cervical cancer, among experienced and less experienced observers, using transvaginal ultrasound (TVS) and magnetic resonance imaging (MRI).MethodsThe TVS observers were all gynecologists and consultant ultrasound specialists, six with and seven without previous experience in cervical cancer imaging. The MRI observers were five radiologists experienced in pelvic MRI and four less experienced radiology residents without previous experience in MRI of the pelvis. The less experienced TVS observers and all MRI observers underwent a short basic training session in the assessment of cervical tumor extension, while the experienced TVS observers received only a written directive. All observers were assigned the same images from cervical cancer patients at all stages (n = 60) and performed offline evaluation to answer the following three questions: (1) Is there a visible primary tumor? (2) Does the tumor infiltrate > ⅓ of the cervical stroma? and (3) Is there parametrial invasion? Interobserver agreement within the four groups of observers was assessed using Fleiss kappa (κ) with 95% CI.ResultsExperienced and less experienced TVS observers, respectively, had moderate interobserver agreement with respect to tumor detection (κ (95% CI), 0.46 (0.40–0.53) and 0.46 (0.41–0.52)), stromal invasion > ⅓ (κ (95% CI), 0.45 (0.38–0.51) and 0.53 (0.40–0.58)) and parametrial invasion (κ (95% CI), 0.57 (0.51–0.64) and 0.44 (0.39–0.50)). Experienced MRI observers had good interobserver agreement with respect to tumor detection (κ (95% CI), 0.70 (0.62–0.78)), while less experienced MRI observers had moderate agreement (κ (95% CI), 0.51 (0.41–0.62)), and both experienced and less experienced MRI observers, respectively, had good interobserver agreement regarding stromal invasion (κ (95% CI), 0.80 (0.72–0.88) and 0.71 (0.61–0.81)) and parametrial invasion (κ (95% CI), 0.69 (0.61–0.77) and 0.71 (0.61–0.81)).ConclusionsWe found interobserver agreement for the assessment of local tumor extension in patients with cervical cancer to be moderate for TVS and moderate‐to‐good for MRI. The level of interobserver agreement was associated with experience among TVS observers only for parametrial invasion. © 2021 The Authors. Ultrasound in Obstetrics & Gynecology published by John Wiley & Sons Ltd on behalf of International Society of Ultrasound in Obstetrics and Gynecology.

International multicenter validation of AI-driven ultrasound detection of ovarian cancer

Abstract Ovarian lesions are common and often incidentally detected. A critical shortage of expert ultrasound examiners has raised concerns of unnecessary interventions and delayed cancer diagnoses. Deep learning has shown promising results in the detection of ovarian cancer in ultrasound images; however, external validation is lacking. In this international multicenter retrospective study, we developed and validated transformer-based neural network models using a comprehensive dataset of 17,119 ultrasound images from 3,652 patients across 20 centers in eight countries. Using a leave-one-center-out cross-validation scheme, for each center in turn, we trained a model using data from the remaining centers. The models demonstrated robust performance across centers, ultrasound systems, histological diagnoses and patient age groups, significantly outperforming both expert and non-expert examiners on all evaluated metrics, namely F1 score, sensitivity, specificity, accuracy, Cohen’s kappa, Matthew’s correlation coefficient, diagnostic odds ratio and Youden’s J statistic. Furthermore, in a retrospective triage simulation, artificial intelligence (AI)-driven diagnostic support reduced referrals to experts by 63% while significantly surpassing the diagnostic performance of the current practice. These results show that transformer-based models exhibit strong generalization and above human expert-level diagnostic accuracy, with the potential to alleviate the shortage of expert ultrasound examiners and improve patient outcomes.

Estimating risk of endometrial malignancy and other intracavitary uterine pathology in women without abnormal uterine bleeding using IETA‐1 multinomial regression model: validation study

ABSTRACTObjectivesTo assess the ability of the International Endometrial Tumor Analysis (IETA)‐1 polynomial regression model to estimate the risk of endometrial cancer (EC) and other intracavitary uterine pathology in women without abnormal uterine bleeding.MethodsThis was a retrospective study, in which we validated the IETA‐1 model on the IETA‐3 study cohort (n = 1745). The IETA‐3 study is a prospective observational multicenter study. It includes women without vaginal bleeding who underwent a standardized transvaginal ultrasound examination in one of seven ultrasound centers between January 2011 and December 2018. The ultrasonography was performed either as part of a routine gynecological examination, during follow‐up of non‐endometrial pathology, in the work‐up before fertility treatment or before treatment for uterine prolapse or ovarian pathology. Ultrasonographic findings were described using IETA terminology and were compared with histology, or with results of clinical and ultrasound follow‐up of at least 1 year if endometrial sampling was not performed. The IETA‐1 model, which was created using data from patients with abnormal uterine bleeding, predicts four histological outcomes: (1) EC or endometrial intraepithelial neoplasia (EIN); (2) endometrial polyp or intracavitary myoma; (3) proliferative or secretory endometrium, endometritis, or endometrial hyperplasia without atypia; and (4) endometrial atrophy. The predictors in the model are age, body mass index and seven ultrasound variables (visibility of the endometrium, endometrial thickness, color score, cysts in the endometrium, non‐uniform echogenicity of the endometrium, presence of a bright edge, presence of a single dominant vessel). We analyzed the discriminative ability of the model (area under the receiver‐operating‐characteristics curve (AUC); polytomous discrimination index (PDI)) and evaluated calibration of its risk estimates (observed/expected ratio).ResultsThe median age of the women in the IETA‐3 cohort was 51 (range, 20–85) years and 51% (887/1745) of the women were postmenopausal. Histology showed EC or EIN in 29 (2%) women, endometrial polyps or intracavitary myomas in 1094 (63%), proliferative or secretory endometrium, endometritis, or hyperplasia without atypia in 144 (8%) and endometrial atrophy in 265 (15%) women. The endometrial sample had insufficient material in five (0.3%) cases. In 208 (12%) women who did not undergo endometrial sampling but were followed up for at least 1 year without clinical or ultrasound signs of endometrial malignancy, the outcome was classified as benign. The IETA‐1 model had an AUC of 0.81 (95% CI, 0.73–0.89, n = 1745) for discrimination between malignant (EC or EIN) and benign endometrium, and the observed/expected ratio for EC or EIN was 0.51 (95% CI, 0.32–0.82). The model was able to categorize the four histological outcomes with considerable accuracy: the PDI of the model was 0.68 (95% CI, 0.62–0.73) (n = 1532). The IETA‐1 model discriminated very well between endometrial atrophy and all other intracavitary uterine conditions, with an AUC of 0.96 (95% CI, 0.95–0.98). Including only patients in whom the endometrium was measurable (n = 1689), the model's AUC was 0.83 (95% CI, 0.75–0.91), compared with 0.62 (95% CI, 0.52–0.73) when using endometrial thickness alone to predict malignancy (difference in AUC, 0.21; 95% CI, 0.08–0.32). In postmenopausal women with measurable endometrial thickness (n = 848), the IETA‐1 model gave an AUC of 0.81 (95% CI, 0.71–0.91), while endometrial thickness alone gave an AUC of 0.70 (95% CI, 0.60–0.81) (difference in AUC, 0.11; 95% CI, 0.01–0.20).ConclusionThe IETA‐1 model discriminates well between benign and malignant conditions in the uterine cavity in patients without abnormal bleeding, but it overestimates the risk of malignancy. It also discriminates well between the four histological outcome categories. © 2023 International Society of Ultrasound in Obstetrics and Gynecology.