分类决策树辅助盆腔MRI术前诊断子宫内膜癌伴卵巢恶性肿瘤

目的采用分类决策树方法,探索盆腔MRI结合患者临床信息在术前诊断子宫内膜癌合并卵巢恶性肿瘤中的价值。方法回顾性分析801例子宫内膜癌患者资料,其中伴卵巢恶性肿瘤者(EC-OMT组)58例,不伴者(EC组)743例,比较两组患者术前临床资料及盆腔MRI征象,计算盆腔MRI对子宫内膜癌合并卵巢恶性肿瘤的诊断效能,采用决策树模型筛选有效的术前诊断指标。结果 EC-OMT组子宫内膜病变浸润肌层深度超过EC组,其侵犯宫颈和宫角、发生腹盆腔淋巴结转移、腹膜转移的比例以及附件区发现肿物的比例均大于EC组,差异均有统计学意义(P均0.01)。两组间宫旁受累差异无统计学意义(1.72%vs 0.40%,P=0.26)。盆腔MRI术前诊断子宫内膜癌合并卵巢恶性肿瘤的敏感度为51.72%(30/58),特异度为99.87%(742/743)。经决策树模型筛选出3项有诊断意义的指标:宫角受累、附件区肿物及血清CA125,诊断子宫内膜癌合并卵巢恶性肿瘤的敏感度为89.66%(52/58)。结论采用分类决策树方法对盆腔MRI和相关临床信息建立决策树模型,可提高对子宫内膜癌合并卵巢恶性肿瘤的术前诊断敏感度。

Decision tree analysis of pelvic MRI in diagnosis of endometrial cancer with ovarian malignant tumor

Objective To explore the value of pelvic MRI combined with clinical information in diagnosis of endometrial cancer (EC) with ovarian malignant tumor (OMT) using decision tree analysis. Methods The clinical information and pelvic MRI characteristics of 58 cases with ovarian malignant tumor (EC-OMT group) and 743 cases without ovarian malignant tumor (EC group) were reviewed and compared. The diagnostic efficacy of pelvic MRI was evaluated. Decision tree analysis was used in determining the performance on the diagnosis. Results In EC-OMT group, the depth of myometrial invasion, the frequency of cervical and cornua uteri involvement, adnexal mass, pelvic or para-aortic lymph nodes involvement and peritoneum metastasis were higher than those in EC group (all P<0.01). Para-uterine involvement showed no significant difference between two groups (1.72% vs 0.40%, P=0.26). In diagnosis of EC with OMT, the sensitivity and specificity value of MRI was 51.72% (30/58) and 99.87% (742/743), respectively. Cornua uteri involvement, adnexal mass and CA125 level were screened as helpful indicators for pre-operation diagnosis by decision tree, and the sensitivity was 89.66% (52/58). Conclusion The diagnosis model of pelvic MRI combined with clinical information by using decision tree analysis can promote sensitivity in diagnosis of EC with OMT.