Value of characteristics of magnetic resonance imaging in predicting the increta type of pernicious placenta previa by using the sequential multivariate logistic regression analysis
张冰,陆芳菲,李茗,陈飞,闫娴,孙艺,胡安宁,朱小倩,陈文倩,王坤
DOI: https://doi.org/10.3760/cma.j.cn101202-20200707-00208
2021-01-01
Abstract:Objective:This study aims to explore the value of image characteristics of magnetic resonance (MR) in predicting the increta type of pernicious placenta previa (PPP).Methods:The clinical and the MR data of 67 patients with PPP and placenta increta in the Affiliated Drum Tower Hospital of Nanjing University Medical School were retrospectively analyzed, aged from 21 years to 48 years (31.85 years±5.12 years) and gestational age ranging from 28 weeks to 38 weeks (32.37 weeks±2.47 weeks). According to the clinical and the pathological diagnostic criteria of different types of PPP with placenta implantation, patients were divided into accreta (25 cases), increta (29 cases), and percreta (13 cases) groups. The clinical data and the MR image characteristics of the three groups of patients were analyzed using the single-factor analysis through the Kruskal-Wallis H test to screen the possible correlation factors of different increta types. The sequential multivariate logistic regression was used for the multivariate analysis of possible correlation factors (MR image characteristics), and the independent risk factors of different increta types were screened. The value of independent risk factors (MR image characteristics) in predicting the increta type of PPP was further analyzed. Results:The single-factor analysis showed that the myometrium interrupt/placenta signal present in the myometrium, localized protuberance of uterus/placenta bulges outward, abnormal placental vessels, inhomogeneous placental signal, enlarged lower part of the uterus with irregular endocervical placenta, and discontinuous low signal bands in the cysto-uterine space were the possible related factors of different increta types (all P values<0.1). Among these factors, discontinuous low-signal bands in the cysto-uterine space ( X1, OR=20.25), myometrium interrupt/placenta signal present in the myometrium ( X2, OR=13.87), localized protuberance of uterus/placenta bulges outward ( X3, OR=6.16), and enlarged lower part of the uterus with irregular endocervical placenta ( X4, OR=4.55) were the independent risk factors of different increta types (all P values<0.05). The regression model logit( P)=1.08/5.54+ 3.01 X1+ 2.63 X2+ 1.82 X3+ 1.52 X4 was established to predict the type of placenta increta. The accuracy of this model in predicting the three types of implantation was 71.64%, and the accuracy in differentiating the group accreta with other groups was 85.07%. Conclusions:MR image characteristics have a high value in predicting the increta type of PPP, showing certain guiding significance for the clinical preoperative evaluation of placenta increta types and selection of different treatment plans.