摘要
目的建立以前列腺特异性抗原(PSA)、MRI及经直肠前列腺超声造影(CE-TRUS)指标为自变量的前列腺癌(PCa)的Logistic回归模型,评价其对PCa的诊断价值。方法选择2017年1月至2018年1月在成都市第三人民医院就诊的223例前列腺疾病患者,年龄44~84岁,平均年龄67.93岁。以病理诊断结果为"金标准",选择有统计学意义的血清PSA及影像学检查指标(MRI、CE-TRUS)为自变量,以病理诊断结果是否为PCa为因变量,建立Logistic回归模型,并通过受试者操作特性(ROC)曲线评估此模型诊断PCa的效能。结果 223例患者中,116例(52.02%)经活组织检查证实为PCa(PCa组),107例(47.98%)为良性前列腺病变(非PCa组)。两组除PSA水平4~10μg/L段差异无统计学意义(P> 0.05)外,其余PSA各段水平差异均具有统计学意义(P <0.05)。CE-TRUS及MRI检查图像阳性特征比较,前列腺的CE-TRUS、MRI异常强化及内外腺边界之间的差异有统计学意义(P <0.05)。结合患者的PSA值、CE-TRUS异常强化、MRI异常强化及内外腺边界的Logistic回归模型预测PCa的准确度91.4%,灵敏度85.1%,特异度92.6%,其ROC曲线下面积为0.914。结论 PSA联合MRI、CE-TRUS等因素建立的Logistic回归模型对PCa预测效果较好,有可能为临床诊断PCa提供可靠的依据。
Objective To establish the Logistic regression model of prostate cancer(PCa) with prostate-specific antigen(PSA),MRI, and contrast enhanced-transrectal ultrasound(CE-TRUS) as independent variables, and evaluate its diagnostic value for PCa. Methods From January 2017 to January 2018, a total of 223 PCa patients were enrolled, aged 44-84 years old with mean age of 67.93 years old. Based on the "gold standard"-pathological diagnosis result, the statistically significant serum PSA and imaging examination indicators(MRI, CE-TRUS) were considered as independent variables, and pathological diagnosis result was the PCa dependent variable, then Logistic regression model was established and the model effectiveness in diagnosing PCa with receiver operating characteristic(ROC) curve were evaluated. Results Among 223 patients, 116(52.02 %)were PCa(PCa group) and 107(47.98 %) were benign prostate disease(non-PCa group) confirmed by biopsy. The difference in various PSA levels of 2 groups were statistically significant(P < 0.05) except for at 4-10 μg/L(P > 0.05). Compared with positive features between CE-TRUS and MRI images, the difference between prostate CE-TRUS, MRI abnormal enhancement, and internal and external gland boundaries were statistically significant(P <0.05). Combined PSA value, CE-TRUS abnormal enhancement, MRI abnormal enhancement, and internal and external gland boundary Logistic regression model, the accuracy of predicting PCa was 91.4 %, sensitivity was 85.1 %, specificity was 92.6 %, and under ROC curve area was 0.914. Conclusion It is demonstrated that Logistic regression model based on PSA combined with MRI, CE-TRUS showed good predictive effect on PCa, which could provide reliable basis for clinical diagnosis of PCa.
作者
徐烽
阮静雅
魏小芳
周洋
XU Feng;RUAN Jing-ya;WEI Xiao-fang;ZHOU Yang(Department of Ultrasound,Chengdu Third People's Hospital,Chengdu 610031,Sichuan,China;Aba Forestry Central Hospital of Sichuan Province,Chengdu 611830,China)
出处
《生物医学工程与临床》
CAS
2020年第6期714-718,共5页
Biomedical Engineering and Clinical Medicine
基金
成都市科技惠民项目(2015-HM01-00129-SF)。
