| Objective: Endometrial polyp is a gynecological disease with a high incidence.Adenomyomatous endometrial polyp is a special pathological type of endometrial polyp,and its clinical manifestations are similar to simple endometrial polyp.At present,the main methods for diagnosing endometrial lesions are ultrasound,diagnostic curettage and hysteroscopy.In published the literature,the diagnosis rate of adenomyomatous endometrial polyp by ultrasound is low,while hysteroscopy and diagnostic curettage are more effective for adenomyomatous endometrial polyp.Although the diagnosis rate of adenomyomatous polyps is high by hysteroscopy and diagnostic curettage,the operation is complicated.And as invasive examinations,their require for the physiological condition of the patients is high.Therefore,it is of great significance to explore a non-invasive examination method with high diagnosis rate.This article will discuss the application value of ultrasound radiomics analysis in distinguishing simple endometrial polyps from adenomyomatous endometrial polyps.Methods:This article included a total of 199 cases of simple endometrial polyps and 140 cases of adenomyomatous endometrial polyps after surgical removal of endometrial polyps in our hospital between 2017.1 and 2019.12.I selected the ultrasound images of the patients,and delineate the region of interest in the largest long-axis cross-sectional image showing the lesion.I used ITK-SNAP software to manually outline the region of interest(ROI)of the primary lesion,and extracted the radiomic features of the ultrasound image.Then the least absolute shrinkage and selection operator(LASSO)were applied to select the relevant radiomic feature tags,and establish a model based on the selected features through logistic regression analysis.Accuracy,specificity,sensitivity and receiver operating characteristic curve(ROC)and the area under curve(AUC)were used to evaluate the ability of the model in identifying two kinds of endometrial polyps.The closer the AUC is to 1,the better the identification ability.Results: A total of 366 ultrasound radioomics effective features were extracted,including 144 gray level co-occurrence matrix,180 gray level run-length matrix features,and 42 histogram features.The 8 most effective features obtained by dimensionality reduction screening are included in the logistic regression model.The model distinguishes simple endometrial polyps and adenomyomatous endometrial polyps in the training set with an AUC of 0.80,a sensitivity of 0.79,a specificity of 0.69,and an accuracy of 0.75;AUC in the test set is 0.81,sensitivity is 0.83,specificity is 0.69,and accuracy is 0.77.Conclusion: The ultrasound-based radiomics model is of medium value in distinguishing simple endometrial polyps from endometrial adenofibroid polyps.Compared with uterine adnexal ultrasound,ultrasonography has improved the ability to distinguish the two kinds of endometrial polyps,but multi-center and more sample further research is still needed. |