首页|期刊导航|磁共振成像|基于胸部CT体成分指标联合前列腺癌病灶与前列腺周围脂肪多参数MRI预测神经侵犯的价值

基于胸部CT体成分指标联合前列腺癌病灶与前列腺周围脂肪多参数MRI预测神经侵犯的价值OA

Predicting perineural invasion in prostate cancer using chest CT body composition and multiparametric MRI of prostate lesions and periprostatic fat

中文摘要英文摘要

目的 探讨基于术前胸部CT体成分指标联合前列腺癌(prostate cancer,PCa)病灶与前列腺周围脂肪组织(periprostatic adipose tissue,PPAT)多参数磁共振成像(multiparametric magnetic resonance imaging,mpMRI)及临床实验室指标预测PCa神经侵犯(perineural invasion,PNI)的价值.材料与方法 回顾性分析75例病理确诊PCa患者,根据术后PNI状态分PNI阳性组(n=35)和PNI阴性组(n=40).基于胸部CT平扫图像,采用Total Segmentator自动获取T12水平肌肉及脂肪面积和比值;mpMRI提取PCa病灶及PPAT的体素内不相干运动(intravoxel incoherent motion,IVIM)及mDIXON-Quant定量参数.两名医师在感兴趣区(region of interest,ROI)内测量取平均值,经组内相关系数(intra-class correlation coefficient,ICC)检验一致性.通过差异分析及多因素logistic回归确定PNI独立预测因子,构建单一及联合预测模型,绘制受试者工作特征(receiver operating characteristic,ROC)曲线并评估曲线下面积(area under the curve,AUC),DeLong检验比较AUC差异,联合模型构建列线图,并通过Hosmer-Lemeshow检验和1000次Bootstrap重采样评估模型校准度和稳定性.结果 PNI阳性组肌肉面积[(86.54±16.42)cm2 vs.(76.52±14.84)cm2]、PPAT-脂肪分数(fat fraction,FF)[(69.37%±9.33%)vs.(62.07%±7.24%)]及IVIM-灌注分数(f)值[0.47(0.33,0.63)vs.0.31(0.25,0.60)]均高于PNI阴性组(P<0.05);红细胞计数(red blood cell count,RBC)低于PNI阴性组[(3.90±0.35)/μL vs.(4.26±0.40)/μL,P<0.05].多因素分析显示肌肉面积[AUC=0.668,95%置信区间(confidence interval,CI):0.545~0.790]、PPAT-FF(AUC=0.724,95%CI:0.610~0.839)及RBC(AUC=0.673,95%CI:0.550~0.796)为独立预测因子.联合模型(T12肌肉面积+PPAT-FF+RBC)的AUC为0.848(95%CI:0.763~0.933),敏感度为71.429%,特异度为80.000%,其预测效能显著优于各单一指标(DeLong检验:联合模型 vs.肌肉面积,P=0.012;联合模型 vs.PPAT-FF,P=0.023;联合模型 vs.RBC,P=0.008).结论 胸部CT体成分参数和PCa病灶及前列腺周围脂肪的mpMRI定量参数及临床指标可有效提升术前预测PCa PNI的能力.

Objective:To evaluate the value of preoperative chest CT-derived body composition indices combined with quantitative multiparametric magnetic resonance imaging(mpMRI)features of prostate cancer(PCa)lesions and periprostatic adipose tissue(PPAT),together with clinical laboratory indicators,for predicting perineural invasion(PNI)in PCa.Materials and Methods:Seventy five patients with pathologically confirmed PCa were retrospectively enrolled and divided into a PNI-positive group(n=35)and a PNI-negative group(n=40).Muscle and adipose tissue areas and their ratios at the T12 vertebral level were automatically quantified from non-contrast chest CT images using Total Segmentator.Quantitative parameters derived from intravoxel incoherent motion(IVIM)and mDIXON-QUANT sequences were extracted from intraprostatic lesions and PPAT on mpMRI.Regions of interest were independently delineated by two radiologists,and mean values were used after interobserver agreement was confirmed by intra-class correlation coefficient.Independent predictors of PNI were identified using univariate analysis followed by multivariate logistic regression.Single-parameter and combined prediction models were constructed and evaluated using receiver operating characteristic(ROC)analysis and area under the curve(AUC).Model performance was compared using the DeLong test,and a nomogram was developed for visualization of the combined model.Model calibration was assessed using the Hosmer-Lemeshow goodness-of-fit test,and internal validation was performed using bootstrap resampling(1000 repetitions).Results:The PNI-positive group exhibited significantly higher muscle area[(86.54±16.42)cm2 vs.(76.52±14.84)cm2],PPAT-fat fraction(FF)[(69.37%±9.33%)vs.(62.07%±7.24%)],and IVIM-perfusion fraction(f)values[0.47(0.33,0.63)vs.0.31(0.25,0.60)]and significantly lower red blood cell(RBC)levels[(3.90±0.35)/μL vs.(4.26±0.40)/μL]compared with the PNI-negative group(all P<0.05).Multivariate analysis identified muscle area[AUC=0.668,95%confidence interval(CI):0.545 to 0.790],PPAT-FF(AUC=0.724,95%CI:0.610 to 0.839),and RBC(AUC=0.673,95%CI:0.550 to 0.796)as independent predictors of PNI.The combined model(muscle area+PPAT-FF+RBC)achieved an AUC of 0.848(95%CI:0.763 to 0.933),with a sensitivity of 71.4%and a specificity of 80.0%.Its predictive performance was significantly superior to that of each single indicator(DeLong test:combined model vs.muscle area,P=0.012;combined model vs.PPAT-FF,P=0.023;combined model vs.RBC,P=0.008).Conclusions:Integration of chest CT-based body composition metrics with mpMRI quantitative features of prostate cancer lesions and periprostatic adipose tissue,together with clinical laboratory indicators,significantly enhances the preoperative prediction of perineural invasion in prostate cancer.

王思齐;张钦和;王秀林;陈丽华;王洪凯;孙祎楠;刘爱连

大连医科大学附属第一医院放射科,大连 116011||辽宁省超极化磁共振专业技术创新中心,大连 116011||医学影像人工智能技术创新中心,大连 116011大连医科大学附属第一医院放射科,大连 116011||辽宁省超极化磁共振专业技术创新中心,大连 116011||医学影像人工智能技术创新中心,大连 116011大连医科大学附属第一医院干细胞临床研究机构,大连 116011大连医科大学附属第一医院放射科,大连 116011||辽宁省超极化磁共振专业技术创新中心,大连 116011||医学影像人工智能技术创新中心,大连 116011医学影像人工智能技术创新中心,大连 116011||大连理工大学医学部,大连 116011大连理工大学医学部,大连 116011大连医科大学附属第一医院放射科,大连 116011||辽宁省超极化磁共振专业技术创新中心,大连 116011||医学影像人工智能技术创新中心,大连 116011

医药卫生

前列腺癌神经侵犯体成分分析定量计算机断层扫描多参数磁共振成像预测模型

prostate cancerperineural invasionquantitative computed tomographybody composition analysismultiparametric magnetic resonance imagingprediction model

《磁共振成像》 2026 (9)

25-33,48,10

辽宁省科技计划联合计划项目(编号:2025-MSLH-199)Joint Research Program of Liaoning Provincial Science and Technology Plan(No.2025-MSLH-199).

10.12015/issn.1674-8034.2026.09.004

评论