Diagnostic value of contrast-enhanced ultrasound in endometrial lesions.

Contrast-enhanced ultrasound (CEUS) can provide both qualitative and quantitative information on endometrial angiogenesis. The CEUS time–intensity curve (TIC) can quantitatively analyse the wash-in and wash-out of contrast agents in endometrial lesions to differentiate between benign and malignant lesions using TIC parameters.

The study enrolled 316 patients who were diagnosed with and treated for endometrial lesions at Beijing Friendship Hospital, Capital Medical University, between February 2017 and May 2019. The patients were divided into two groups: a benign endometrial lesion group and a malignant endometrial lesion group. The inclusion criteria for benign lesion group are cases with complete data, cases with pathological results of endometrial hyperplasia or polyps. The exclusion criteria for benign lesion group are cases with incomplete data, cases without pathological results, cases with the submucosal myoma, endometrial intraepithelial neoplasia, or uncertain lesions, cases received medication treatment and hysteroscopy treatment. The inclusion criteria for malignant lesion group are cases with complete data, cases with pathological results of endometrial carcinoma. The exclusion criteria for malignant lesion group are cases with incomplete data, cases without pathological results, cases with the pathological result shows atypical hyperplasia of the endometrium, cases received medication treatment and hysteroscopy treatment. The study was conducted in accordance with the Declaration of Helsinki and approved by the Ethics Committee of Beijing Friendship Hospital. A LOGIQ E9 ultrasound system (GE Healthcare, Wauwatosa, WI, USA) using an IC5-9-B (5–9 MHz) transducer was used in this study. The mechanical index was kept at 0.10–0.15. The contrast agent used was SonoVue (Bracco, Milan, Italy), which consists of sulphur hexafluoride gas microbubbles. SonoVue (59 mg) was mixed with 5 mL of 0.9% saline, and a bolus of 2.4 mL was injected into the median cubital vein, followed by a 5–10 mL saline flush. All patients were placed in the lithotomy position after bladder emptying. Transvaginal ultrasound using 2D ultrasonography was conducted to determine the thickness, echo, morphology, and blood flow of the endometrium. After the SonoVue injection, the CEUS mode began recording the enhancement process at the plane where we suspected endometrial pathological areas to be located. The enhancement process took 2–6 min, and the images and videos were stored in the ultrasound system for further analysis. SonoLiver (Image-arena v4.0, TomTec Imaging System, Munich, Germany) was used to analyse the contrast images. We imported contrast movies into the software to obtain the time–intensity curve (TIC) for analysis, and the endometrial lesion area was selected as the region of interest (ROI), where the normal myometrium of the same size at the same depth was selected as the reference object (Fig.  1 ). Then, we collected the intensity parameters according to the SonoLiver analytic software, including intensity ratio of maximum echo (Imax), time to peak (TTP), rise time (RT), mean transit time (mTT), 50% rise slope (Rs50), 10%–90% rise slope (Rs1090), fall time (FT), 50% fall slope (Fs50), fall half time (FHT), area under the curve (AUC) of the receiver operating characteristic curve (ROC), area under the wash-in curve (WinAUC), area under the wash-out curve (WoutAUC), area under the wash-in and wash-out curve (WioAUC), wash-in rate (WinR = WinAUC/RT), wash-out rate (WoutR = WoutAUC/FT), and quality of fit (QOF). The detailed parameters are shown in Fig.  2 , using the green curve of the analysis model as an example. Fig. 1 ROI and reference object selection. The left image is the 2D ultrasound image of thickened endometrium, and the right image is the CEUS image. The green area represents the ROI, the yellow area represents the reference object selection, and the blue area represents the overall image analysis area Fig. 2 TIC and TIC parameters. The yellow curve is the reference model, and the green curve is the analysis model. The peak is the highest point of the analysis model curve, WinAUC is the area under the wash-in curve, WoutAUC is the area under the wash-out curve, WioAUC is the area under the wash-in and wash-out curve. TTP is the time to peak, mTT is the mean transit time, RT is the rise time, and FT is the fall time. Ti is the intersection point of the rise slope of the wash-in curve and the time axis, and To is the intersection point of the fall slope of the wash-out curve and the time axis ROI and reference object selection. The left image is the 2D ultrasound image of thickened endometrium, and the right image is the CEUS image. The green area represents the ROI, the yellow area represents the reference object selection, and the blue area represents the overall image analysis area TIC and TIC parameters. The yellow curve is the reference model, and the green curve is the analysis model. The peak is the highest point of the analysis model curve, WinAUC is the area under the wash-in curve, WoutAUC is the area under the wash-out curve, WioAUC is the area under the wash-in and wash-out curve. TTP is the time to peak, mTT is the mean transit time, RT is the rise time, and FT is the fall time. Ti is the intersection point of the rise slope of the wash-in curve and the time axis, and To is the intersection point of the fall slope of the wash-out curve and the time axis Imax is the intensity ratio of maximum echo, TTP is the time to peak of the TIC, RT is the rise time of the TIC, mTT is the mean transit time of the TIC, Rs50 is the rise slope to 50% of the TIC, Rs1090 is the rise slope from 10 to 50% of the TIC, FT is the fall time of the TIC, Fs50 is the fall slope to 50% of the TIC, FHT is the time of the TIC fall to the half, AUC is the area under the curve of the TIC, WinAUC is the area under the wash-in curve which is the AUC from Ti to peak, WoutAUC is the area under the wash-out curve which is the AUC from peak to To, WioAUC is area under the wash-in and wash-out curve which is the AUC from Ti to To, WinR is the filling rate of wash-in period, WoutR is the evacuation rate of wash-out period, QOF is quality of fit of TIC which is the similarity of the TIC and the actual change, the value of QOF is better close to 100% and the acceptable range is above 68%. Statistical analyses were performed using SPSS software (version 23.0; IBM, Chicago, United States). Data were expresses as mean ± standard deviation. A non-parametric test was used for parameter comparison. The cut-off values for diagnosis were obtained from the ROC curve. The model was built using logistic regression. Statistical significance was set at P  < 0.05.

The TIC formed by the SonoLiver analysis software is shown in Fig.  3 . The QOF for all cases was greater than 80%, with a satisfactory QOF between the compressed signal and the bolus perfusion model. Fig. 3 TIC of endometrial lesions. A is the TIC of benign lesions, and B is the TIC of malignant lesions TIC of endometrial lesions. A is the TIC of benign lesions, and B is the TIC of malignant lesions There are a total 316 cases with endometrial lesions underwent CEUS examination. Among them, there are 36 cases of malignant lesions, including 32 cases with Endometrioid carcinoma and 4 cases without pathological results, 5 cases with EIN, 275 cases of benign lesions, including 204 cases of endometrial hyperplasia or polyps, 15 cases of benign submucosal fibroids, 56 cases without pathological results. By inclusion and exclusion criteria, a total of 236 patients were included in this study: 204 patients with benign lesions (mean age, 49.22 ± 11.84 [range, 22–81] years) and 32 patients with malignant lesions (mean age, 58.72 ± 11.45 [range, 19–79] years). There are 118 postmenopausal patients and 118 premenopausal patients in the benign lesions group, and 28 postmenopausal patients and 4 premenopausal patients in the malignant lesions group. Among the cases of benign lesions, 62 cases were of endometrial hyperplasia and 142 of endometrial polyps. The difference in age between the two groups was statistically significant ( P  < 0.001). The endometrial thickness of the benign lesion group was 10.03 ± 4.44 [range, 2.00–32.00]mm and the malignant lesion group was 17.38 ± 12.19 [range, 5.00–49.00]mm, the difference in endometrial thickness between the two groups was statistically significant ( P  < 0.001). TIC parameters of the benign and malignant groups are shown in Table  1 . The malignant lesion group had higher Imax, Rs50, and Rs1090; lower Fs50; shorter TTP, RT, FT, FHT, and mTT; and larger AUC, WinAUC, WoutAUC, and WioAUC than the benign lesion group. Table 1 Comparison of TIC parameters between benign and malignant endometrial lesions Benign lesions Malignant lesions P -value Imax 89.31 ± 136.55 908.42 ± 1285.17 < 0.001 TTP 14.56 ± 5.95 9.22 ± 2.31 < 0.001 RT 10.08 ± 4.13 6.89 ± 1.76 < 0.001 Rs50 23.93 ± 39.66 162.80 ± 314.29 < 0.001 Rs1090 20.32 ± 34.44 137.77 ± 252.93 < 0.001 FT 11.53 ± 5.64 8.92 ± 3.37 0.005 Fs50 −4.95 ± 10.37 −28.88 ± 56.72 < 0.001 FHT 29.49 ± 11.74 22.71 ± 6.52 0.001 mTT 31.04 ± 19.96 21.06 ± 8.12 < 0.001 AUC 19,648.65 ± 27,060.70 140,506.83 ± 274,972.69 0.001 WinAUC 1662.97 ± 2618.98 9903.64 ± 18,631.20 < 0.001 WoutAUC 6333.78 ± 9710.43 40,918.85 ± 81,368.31 < 0.001 WioAUC 6454.92 ± 9965.88 42,136.68 ± 83,294.15 < 0.001 WinR 181.56 ± 264.28 1419.16 ± 2128.24 < 0.001 WoutR 591.13 ± 734.45 4812.31 ± 8529.87 < 0.001 Imax Intensity ratio of maximum echo, TTP Time to peak, RT Rise time, Rs50 50% rise slope, Rs1090 10%–90% rising slope, FT Fall time, Fs50 50% fall slope, FHT Fall half time, mTT Mean transit time, AUC Area under the curve of the receiver operating characteristic curve, WinAUC Area under the wash-in curve, WoutAUC Area under the wash-out curve, WioAUC Area under the wash-in and wash-out curve, WinR = WinAUC/RT Wash-in rate, WoutR = WoutAUC/FT Wash-out rate Comparison of TIC parameters between benign and malignant endometrial lesions Imax Intensity ratio of maximum echo, TTP Time to peak, RT Rise time, Rs50 50% rise slope, Rs1090 10%–90% rising slope, FT Fall time, Fs50 50% fall slope, FHT Fall half time, mTT Mean transit time, AUC Area under the curve of the receiver operating characteristic curve, WinAUC Area under the wash-in curve, WoutAUC Area under the wash-out curve, WioAUC Area under the wash-in and wash-out curve, WinR = WinAUC/RT Wash-in rate, WoutR = WoutAUC/FT Wash-out rate The ROC of the TIC parameters is shown in Fig.  4 , and the sensitivity, specificity, and cut-off values of the parameters are shown in Table  2 . As the values of Imax, Rs50, Rs1090, AUC, WinAUC, WoutAUC, WioAUC, and age increased, the tumours tended to be malignant. By contrast, when TTP, RT, FT, FHT, mTT, and Fs50 increased, the tumours tended to be benign. According to the curves, Imax, Rs50, and Rs1090 showed good sensitivity; however, the specificities of Rs50 and Rs1090 were < 70%. The specificities of FHT, AUC, and WoutAUC were higher than 80%, but their sensitivities were lower than 50%. None of the parameters had good sensitivity and specificity simultaneously; therefore, logistic regression was continued. Fig. 4 ROC curve of TIC parameters. A is the ROC curve of TTP, RT, FT, FHT, mTT, and Fs50. B is the ROC curve of Imax, Rs50, Rs1090, AUC, WinAUC, WoutAUC, and WioAUC. Imax: intensity ratio of maximum echo; TTP: time to peak; RT: rise time; FT: fall time; Rs50: 50% rise slope; Rs1090: 10%–90% rising slope; Fs50: 50% fall slope; FHT: fall half time; mTT: mean transit time; AUC: area under the curve of the receiver operating characteristic curve; WinAUC: area under the wash-in curve; WoutAUC: area under the wash-out curve; WioAUC: area under the wash-in and wash-out curve; WinR = WinAUC/RT: wash-in rate; WoutR = WoutAUC/FT: wash-out rate Table 2 ROC curve results of parameter analysis of TIC for endometrial lesions ROC-AUC Sensitivity (%) Specificity (%) Cut-off Imax 0.912 (0.857–0.966) 96.9 74.5 97.08 TTP 0.833 (0.765–0.901) 74.5 81.2 10.8 RT 0.789 (0.713–0.864) 74.5 75.0 7.51 FT 0.833 (0.765–0.901) 74.5 81.2 10.8 Rs50 0.796 (0.724–0.869) 90.6 62.7 13.82 Rs1090 0.802 (0.730–0.874) 93.8 60.8 10.63 Fs50 0.781 (0.694–0.867) 71.1 78.1 −4.32 FHT 0.684 (0.593–0.774) 48.5 81.2 27.98 mTT 0.704 (0.612–0.797) 67.6 75.0 22.72 AUC 0.677 (0.575–0779) 37.5 89.2 44,541.86 WinAUC 0.731 (0.634–0.0.828) 71.9 64.2 1444.10 WoutAUC 0.696 (0.597–0.796) 46.9 82.8 9731.43 WioAUC 0.706 (0.607–0.804) 68.8 62.7 4636.79 WinR 0.848 (0.772–0.924) 78.1 78.4 266.60 WoutR 0.739 (0.645–0.833) 71.9 70.1 595.74 Age 0.752 (0.662–0.842) 78.1 71.6 54.5 Imax Intensity ratio of maximum echo, TTP Time to peak, RT Rise time, FT Fall time, Rs50 50% rise slope, Rs1090 10%–90% rising slope, Fs50 50% fall slope, FHT Fall half time, mTT Mean transit time, AUC Area under the curve of the receiver operating characteristic curve, WinAUC Area under the wash-in curve, WoutAUC Area under the wash-out curve, WioAUC Area under the wash-in and wash-out curve, WinR = WinAUC/RT Wash-in rate, WoutR = WoutAUC/FT Wash-out rate ROC curve of TIC parameters. A is the ROC curve of TTP, RT, FT, FHT, mTT, and Fs50. B is the ROC curve of Imax, Rs50, Rs1090, AUC, WinAUC, WoutAUC, and WioAUC. Imax: intensity ratio of maximum echo; TTP: time to peak; RT: rise time; FT: fall time; Rs50: 50% rise slope; Rs1090: 10%–90% rising slope; Fs50: 50% fall slope; FHT: fall half time; mTT: mean transit time; AUC: area under the curve of the receiver operating characteristic curve; WinAUC: area under the wash-in curve; WoutAUC: area under the wash-out curve; WioAUC: area under the wash-in and wash-out curve; WinR = WinAUC/RT: wash-in rate; WoutR = WoutAUC/FT: wash-out rate ROC curve results of parameter analysis of TIC for endometrial lesions Imax Intensity ratio of maximum echo, TTP Time to peak, RT Rise time, FT Fall time, Rs50 50% rise slope, Rs1090 10%–90% rising slope, Fs50 50% fall slope, FHT Fall half time, mTT Mean transit time, AUC Area under the curve of the receiver operating characteristic curve, WinAUC Area under the wash-in curve, WoutAUC Area under the wash-out curve, WioAUC Area under the wash-in and wash-out curve, WinR = WinAUC/RT Wash-in rate, WoutR = WoutAUC/FT Wash-out rate Association between the dichotomous dependent variables and continuous variables was assessed using the binary logistic regression model. We analysed the common parameters and those with high sensitivity or specificity to obtain the model. Imax, TTP, RT, Rs50, Rs1090, WinAUC, WiAUC, WinR, WoutR, and age were used as independent variables and endometrial pathological results were used as dependent variables for binary logistic regression. The missing data are deleted and retain complete data for analysis. We use forward: Wald stepwise into the binary logistic regression, which involves screening independent variables step by step to solve the problem of multicollinearity between variables. After model fitting, the final result of regression analysis showed that Imax, TTP, WinR, and age were related to the diagnosis of benign and malignant endometrial lesions using CEUS (χ 2  = 94.914, P  < 0.001). The logistic regression model was as follows: In(benign/malignant) = −4.645 + 0.007Imax—0.549ttp + 0.119age + 0.002WinR. The ROC curve for this model is shown in Fig.  5 . The AUC of the ROC curve was 0.969 (0.940–0.999). The sensitivity and specificity of the model were 93.8% and 92.2%, respectively. Fig. 5 ROC curve of logistic regression of parameters with Imax, TTP, WinR, and age. Imax: intensity ratio of maximum echo; TTP: time to peak; WinR = WinAUC/RT: area under the wash-in curve/rise time ROC curve of logistic regression of parameters with Imax, TTP, WinR, and age. Imax: intensity ratio of maximum echo; TTP: time to peak; WinR = WinAUC/RT: area under the wash-in curve/rise time

Endometrial diseases are common gynaecological diseases affecting women’s health, and the incidence of endometrial cancer is increasing globally [ 1 ]. Ultrasonography is a common examination method, has increasingly being applied in various clinical fields [ 2 , 3 ]. However, 2D examinations have certain limitations in the diagnosis of benign and malignant endometrial lesions [ 4 ]. Contrast-enhanced ultrasound (CEUS) has been widely used in disease diagnosis in various fields in recent years and can evaluate and quantify microcirculation in a particular area in real time [ 5 , 6 ]. CEUS is widely used in the diagnosis of hepatic diseases, as well as in breast, cardiovascular, urinary, and abdominal interventional therapies [ 7 – 10 ]. However, CEUS has not been widely applied in the field of gynaecology. In previous studies, researches are mainly limited to distinguishing between benign and malignant adnexal masses [ 11 , 12 ], evaluate the staging of cervical carcinoma and evaluate the therapeutic effect of adenomyosis [ 13 , 14 ]. While the current studies on CEUS of endometrial mainly focus on qualitative studies, in quantitative studies, there are relatively few contrast imaging parameters included [ 6 , 15 , 16 ]. In our study, we aim to search more quantitative parameters to evaluate benign and malignant endometrial lesions, and analyse and establish a predictive model for endometrial through multi parameters of CEUS, further improving the accuracy of diagnosis of endometrial lesions.

Endometrial diseases are common gynaecological diseases; benign lesions usually cause irregular vaginal bleeding, increased and decreased menstrual volume, and infertility, and some benign lesions might progress to malignancy [ 1 ]. However, differentiating between benign and malignant endometrial lesions using 2D ultrasonography is sometimes difficult [ 4 ]. CEUS has developed rapidly in recent years and can accurately detect vascular conditions and blood perfusion in lesions. CEUS has been widely used in disease diagnosis and function evaluation in other fields. However, its application in the diagnosis of uterine diseases remains rare [ 10 ]. In our study, the absolute values of all TIC parameters differed significantly between the benign and malignant lesion groups. Compared with benign lesions, malignant lesions had a higher Imax, faster peak and decline, larger rising slope, smaller falling slope, larger area under the perfusion curve, and higher wash-in and wash-out times, generally showing a steeper inclining curve, which then declined rapidly. This enhancement mode can be described as “fast wash-in, fast wash-out and high enhancement”. The curve for benign lesions was smoother, the peak was rounder and blunt, and the rise and fall were slower. This is the most obvious difference between benign and malignant lesions. This is mainly related to the pathological changes of endometrial carcinoma. In malignant lesions, the tumours have a large amount of anarchic neovascularisation without the anterior sphincter of capillaries, resulting in low blood flow resistance [ 3 , 17 , 18 ]. At the same time,there is distortion of new blood vessels and formation of arteriovenous fistulas in the lesions [ 16 ]. This is the reason for the formation of different CEUS characteristics in benign and malignant endometrial lesions. In the ROC curve analysis of this study, all CEUS parameters were included to determine the diagnostic ability. In all of the CEUS parameters in our study, only Imax with an AUC of the ROC curve > 0.90 had good diagnostic performance, whereas FHT with the AUC of the ROC curve < 0.70 and other parameters with an AUC of the ROC curve of 0.70–0.90 had medium diagnostic performance. However, Imax had high sensitivity but low specificity, whereas the other parameters did not have high sensitivity or specificity at the same time. Therefore, we used logistic regression to screen the TIC parameters to obtain a model with better diagnostic efficiency [ 16 ]. Although in this study all the TIC parameters had statistical significance in distinguishing between benign and malignant lesions, there were numerous parameters that could not be used for effective logistic regression. Therefore, it was necessary to select the most representative parameters that could best predict benign and malignant endometrial lesions for modelling. Accordingly, in combination with literature reports, clinical practice, and our ROC curve results, we analysed Imax, TTP, RT, Rs50, Rs1090, WinAUC, WiAUC, Win R, Wout R, and age as independent variables and finally selected four parameters—Imax, TTP, WinR, and age—as the main variables to evaluate endometrial lesions diagnosed using CEUS. Imax could better reflect the high echo intensity of malignant endometrial lesions by CEUS, whereas TTP showed that it reached its peak faster. These results are similar to previous studies [ 4 , 16 , 19 ]. In previous studies, the incidence rate of endometrial carcinoma increased with age [ 20 ]. WinR is a relatively novel TIC parameter used in this study and reflects the perfusion rate of the TIC during the perfusion period, that is, the ratio of WinAUC to RT, which represents the contrast ability of the lesion during the perfusion period. The WinR of malignant lesions was significantly higher than that of benign lesions. The use of multivariate analysis to build the prediction model can create multiple indicator variable combinations, which are more predictable and accurate than a single indicator variable. In this study, the sensitivity and specificity of the model, built using logistic regression, were 93.8% and 92.2%, respectively, which were higher than those of the single variable models and had better diagnostic efficacy. CEUS can not only provide vascular information about endometrial polyps, but also diagnose endometrial carcinoma in the early stage, observe the extent of myometrial invasion, and the extent of uterine tissue invasion, providing useful information for surgeons [ 3 , 9 ]. Although this was able to obtain better contrast-enhanced endometrial imaging information and estimate benign and malignant lesions, it had some limitations. FIGO staging of endometrioid carcinoma was not performed in malignant group, we would compare endometrial lesions of FIGO stage I with benign lesions in order to improve the diagnostic value of CEUS for endometrial carcinoma in the future study. This study included a large number of benign cases and a relatively small number of malignant cases. Due to the small number of malignant cases, only a few variables were included in the logistic regression analysis. In follow-up studies, more malignant cases can be collected, and the logistic regression model can be improved, quantitative analysis will also be conducted on different pathological types of endometrial carcinoma to improve the diagnostic value of CEUS for different subtypes. This study used only 2D ultrasound images combined with CEUS to diagnose lesions. The overall evaluation of the lesions had certain limitations that may have led to smaller malignant lesions that could not be observed with CEUS. In a follow-up study, three-dimensional imaging can be combined to comprehensively evaluate the disease and improve the diagnostic accuracy. CEUS can provide both qualitative and quantitative information on endometrial angiogenesis. CEUS TIC can quantitatively analyse the wash-in and wash-out of contrast agents in endometrial lesions to differentiate between benign and malignant lesions using TIC parameters. Therefore, it can provide a new method for preoperative differentiation of benign and malignant endometrial lesions.