Guorong Lyu

Development of a High‐Performance Ultrasound Prediction Model for the Diagnosis of Endometrial Cancer

Objectives To develop and validate an ultrasonography‐based machine learning (ML) model for predicting malignant endometrial and cavitary lesions. Methods This retrospective study was conducted on patients with pathologically confirmed results following transvaginal or transrectal ultrasound from 2021 to 2023. Endometrial ultrasound features were characterized using the International Endometrial Tumor Analysis (IETA) terminology. The dataset was ranomly divided (7:3) into training and validation sets. LASSO (least absolute shrinkage and selection operator) regression was applied for feature selection, and an extreme gradient boosting (XGBoost) model was developed. Performance was assessed via receiver operating characteristic (ROC) analysis, calibration, decision curve analysis, sensitivity, specificity, and accuracy. Results Among 1080 patients, 6 had a non‐measurable endometrium. Of the remaining 1074 cases, 641 were premenopausal and 433 postmenopausal. Performance of the XGBoost model on the test set: The area under the curve (AUC) for the premenopausal group was 0.845 (0.781–0.909), with a relatively low sensitivity (0.588, 0.442–0.722) and a relatively high specificity (0.923, 0.863–0.959); the AUC for the postmenopausal group was 0.968 (0.944–0.992), with both sensitivity (0.895, 0.778–0.956) and specificity (0.931, 0.839–0.974) being relatively high. SHapley Additive exPlanations (SHAP) analysis identified key predictors: endometrial–myometrial junction, endometrial thickness, endometrial echogenicity, color Doppler flow score, and vascular pattern in premenopausal women; endometrial thickness, endometrial–myometrial junction, endometrial echogenicity, and color Doppler flow score in postmenopausal women. Conclusion The XGBoost‐based model exhibited excellent predictive performance, particularly in postmenopausal patients. SHAP analysis further enhances interpretability by identifying key ultrasonographic predictors of malignancy.

Can the ultrasound microcystic pattern accurately predict borderline ovarian tumors?

Abstract Objective To investigate whether the ultrasound microcystic pattern (MCP) can accurately predict borderline ovarian tumors (BOTs). Methods A retrospective collection of 393 patients who met the inclusion criteria was used as the study population. Indicators that could well identify BOT in different pathological types of tumors were derived by multivariate unordered logistic regression analysis. Finally, the correlation between ultrasound MCP and pathological features was analyzed. Results (1) MCP was present in 55 of 393 ovarian tumors, including 34 BOTs (34/68, 50.0%), 16 malignant tumors (16/88, 18.2%), and 5 benign tumors (5/237, 2.1%). (2) Univariate screening showed significant differences (P < 0.05) in patient age, CA-125 level, ascites, > 10 cyst locules, a solid component, blood flow, and MCP among BOTs, benign ovarian tumors, and malignant ovarian tumors. (3) Multivariate unordered logistic regression analysis showed that the blood flow, > 10 cyst locules, and MCP were significant factors in identifying BOTs (P < 0.05). (4) The pathology of ovarian tumors with MCP showed "bubble"- or "fork"- like loose tissue structures. Conclusion MCP can be observed in different pathological types of ovarian tumors and can be used as a novel sonographic marker to differentiate between BOTs, benign tumors and malignant tumors. MCP may arise as a result of anechoic cystic fluid filling the loose tissue gap.

Microcystic pattern and shadowing are independent predictors of ovarian borderline tumors and cystadenofibromas in ultrasound

To determine the sonographic characteristics of borderline tumors (BoTs) and cystadenofibromas (CAFs). Preoperative sonograms from consecutive patients who had at least one primary epithelial tumor in the adnexa were retrospectively collected. All tumors were described using the International Ovarian Tumor Analysis terminology. Ultrasound variables were tested using multinomial logistic regression after univariate analysis. A total of 650 patients were included in this study. Of these, 110 had a CAF, 128 had a BoT, 249 had a cystadenoma (CAD), and 163 had a cystadenocarcinoma (CAC). Nearly half of CAFs and more than half of BoTs and CACs appeared to be unilocular and multilocular solid on the ultrasound images, while CADs were predominantly uni- or multilocular (p < 0.001). Overall, shadowing was identified in 82/650 cases. Sixty-five of 110 (59.1%) CAFs exhibited an acoustic shadow, compared with only 4/249 (1.6%) in CADs, 7/128 (5.5%) in BoTs, and 6/163 (3.7%) in CACs (p < 0.001). Furthermore, 112/650 cases demonstrated microcystic pattern (MCP). Sixty-eight of 128 (53.1%) BoTs exhibited MCP, compared with only 5/249 (2.0%) in CADs, 19/163 (11.7%) in CACs, and 20/110 (18.2%) in CAFs (p < 0.001). Logistic regression analysis revealed that shadowing is an independent predictor of CAFs, while MCP is an independent predictor of BoTs. Sonographic findings for CAFs and BoTs were complex and partly overlapped with those for CACs. However, proper recognition and utilization of shadowing or MCP may help to correctly discriminate CAFs and BoTs. • Sonographic findings for borderline tumors and cystadenofibromas are complex and mimic malignancy. • Microcystic pattern and shadowing are independent predictors of borderline tumors and cystadenofibromas respectively.

Comparison of O‐RADS, GI‐RADS, and ADNEX for Diagnosis of Adnexal Masses: An External Validation Study Conducted by Junior Sonologists

ObjectiveTo externally validate the Ovarian‐adnexal Reporting and Data System (O‐RADS) and evaluate its performance in differentiating benign from malignant adnexal masses (AMs) compared with the Gynecologic Imaging Reporting and Data System (GI‐RADS) and Assessment of Different NEoplasias in the adneXa (ADNEX).MethodsA retrospective analysis was performed on 734 cases from the Second Affiliated Hospital of Fujian Medical University. All patients underwent transvaginal or transabdominal ultrasound examination. Pathological diagnoses were obtained for all the included AMs. O‐RADS, GI‐RADS, and ADNEX were used to evaluate AMs by two sonologists, and the diagnostic efficacy of the three systems was analyzed and compared using pathology as the gold standard. We used the kappa index to evaluate the inter‐reviewer agreement (IRA).ResultsA total of 734 AMs, including 564 benign masses, 69 borderline masses, and 101 malignant masses were included in this study. O‐RADS (0.88) and GI‐RADS (0.90) had lower sensitivity than ADNEX (0.95) (P &lt; .05), and the PPV of O‐RADS (0.98) was higher than that of ADNEX (0.96) (P &lt; .05). These three systems showed good IRA.ConclusionO‐RADS, GI‐RADS, and ADNEX showed little difference in diagnostic performance among resident sonologists. These three systems have their own characteristics and can be selected according to the type of center, access to patients' clinical data, or personal comfort.