基于超声造影纹理特征联合临床指标预测高危前列腺癌的价值研究

Research on the value of predicting high-risk prostate cancer based on contrast-enhanced ultrasound texture features combined with clinical indicators

目的:构建基于超声造影纹理特征联合临床指标预测高危前列腺癌(PCa)模型,并评价该模型术前预测高危PCa的性能力。方法:回顾性分析2018年2月至2021年10月广西医科大学第一附属医院收治的PCa患者188例,血清前列腺特异性抗原(PSA)均大于4.00 ng/mL,根据病理结果分为高危组(GS>7分)和非高危组(GS≤7分),并随机分为训练队列(n=132)和验证队列(n=56)。从患者的超声造影图像中提取影像组学特征,采用最小绝对收缩与选择算子(LASSO)回归和十折交叉检验进行降维,筛选特征,计算影像组学评分(Radscore),并联合临床指标构建预测模型。采用受试者工作特征(ROC)曲线、校准曲线和决策曲线评价该模型的预测效能。结果:经过LASSO回归及十折交叉检验,279个图像纹理特征降维至5个。Radscore在训练队列和验证队列中评估高危PCa的曲线下面积(AUC)分别为0.75(95%CI:0.67~0.83)和0.76(95%CI:0.63~0.85)。多因素Logistic回归分析显示,PSA密度(PSAD)和Radscore为高危PCa的独立预测因素(P<0.05)。预测模型在训练队列中的AUC为0.77(95%CI:0.69~0.85),在验证队列中的AUC为0.78(95%CI:0.65~0.91)。校准曲线和决策曲线分析结果具有良好的准确性和临床应用价值。结论:基于超声造影纹理特征联合临床指标构建的模型可有效预测高危PCa,该模型可用于术前无创性评估高危PCa,有助于指导临床决策。

Objective:To construct a model for predicting high-risk prostate cancer(PCa)based on contrast-enhanced ultrasound texture features combined with clinical indicators and evaluate its ability to predict high-risk PCa before surgery.Methods:A total of 188 patients with serum prostate specific antigen(PSA)more than4.00ng/ml and pathologically confirmed PCa admitted to the First Affiliated Hospital of Guangxi Medical University from February 2018 to October 2021 were retrospectively analyzed.The patients who were divided into highrisk group(GS>7 points)and non high-risk group(GS≤7 points)according to the pathological results were randomly divided into training cohort(n=132)and validation cohort(n=56).Radiomics features were extracted from contrast-enhanced ultrasound(CEUS)images of patients.The least absolute shrinkage and selection operator(LASSO)regression as well as ten-fold cross test were used to reduce the dimension,screen the features,calculate the radiomics score(Radscore)and construct a prediction model combined with clinical indicators.Receiver operating characteristic(ROC)curve,calibration curve and decision curve were used to evaluate the predictive performance of the model.Results:After LASSO regression and ten-fold cross test,the dimension of 279 image texture features was reduced to 5.The area under the curve(AUC)in the high risk PCa assessed by Radscore in the training and validation cohort were respectively 0.75(95%CI:0.67-0.83)and 0.76(95%CI:0.63-0.85).Multi-variate logistic regression analysis showed that PSA density(PSAD)and Radscore were independent pre-dictors of high-risk PCa(P<0.05).The AUC of the prediction model was 0.77(95%CI:0.69-0.85)in the training cohort and 0.78(95%CI:0.65-0.91)in the validation cohort.The results obtained from the analysis of calibration curve and decision curve had good accuracy and clinical application value.Conclusion:The constructed model based on contrast-enhanced ultrasound texture features combined with clinical indicators can effectively predict high-risk PCa.The model can be used to evaluate high-risk PCa noninvasiveness before surgery,which is helpful to guide clinical decision-making.