分享:
分享到微信朋友圈
X
临床研究
基于Bp-MRI影像组学预测前列腺病变良恶性的效能及风险评估
赵莹莹 方陈 吴声连 徐伟 郑鹏翔 郑伟龙 陈志强

Cite this article as: Zhao YY, Fang C, Wu SL, et al. Prediction and risk assessment of benign and malignant prostate lesions based on Bp-MRI radiomics[J]. Chin J Magn Reson Imaging, 2022, 13(8): 43-47.本文引用格式:赵莹莹, 方陈, 吴声连, 等. 基于Bp-MRI影像组学预测前列腺病变良恶性的效能及风险评估[J]. 磁共振成像, 2022, 13(8): 43-47. DOI:10.12015/issn.1674-8034.2022.08.008.


[摘要] 目的 探讨基于双参数磁共振成像(biparameter magnetic resonance imaging, Bp-MRI)影像组学及临床信息对前列腺良恶性病变的诊断、鉴别及风险评估。材料与方法 回顾性分析161例经病理学证实的前列腺疾病患者病例,按7∶3的比例随机、分层分为训练集和验证集。采用t检验/Wilcoxon秩和检验、最小绝对收缩和选择算子(least absolute shrinkage and selection operator, LASSO)算法、Spearman相关分析和logistic回归模型对临床特征和影像组学特征进行分析,构建影像组学模型及联合模型。通过绘制受试者工作特征(receiver operating characteristic, ROC)曲线并计算曲线下面积(area under the curve, AUC),对模型的性能进行评价。然后,利用放射组学特征和临床特征构建联合列线图,并进行验证。结果 影像.组学模型在训练集和验证集预测前列腺病变良恶性的AUC分别为0.946(95% CI:0.903~0.982)、0.902(95% CI:0.862~0.958);联合模型在两组间预测前列腺病变良恶性的AUC分别为0.965(95% CI:0.904~0.989)、0.924(95% CI:0.868~0.980)。结论 基于Bp-MRI的联合模型对于前列腺癌具有较高的诊断效能。结合总前列腺特异性抗原(total prostate specific antigen, tPSA)、游离前列腺特异抗原(free prostate specific antigen, fPSA)/tPSA比值(f/t)和影像组学特征的联合列线图可能为前列腺疾病患者的风险预测和个体化治疗提供有效工具。
[Abstract] Objective To explore the diagnosis,differential diagnosis and risk assessment of benign and malignant prostatic lesions based on biparameter magnetic resonance imaging (Bp-MRI) radiomics and clinical information.Materials and Methods A total of 161 patient cases with pathologically proven prostate disease were retrospectively analyzed and randomly divided into training set and verification set in 7∶3 ratio. The t-test /Wilcoxon rank sum test, the least absolute shrinkage and selection operator (LASSO) algorithm, Spearman correlation analysis, and logistic regression model were used to analyze the clinical and radiographic features, and the radiographic model and the joint model were constructed. The performance of the model was evaluated by plotting the receiver operating characteristic (ROC) curve and calculating the area under the curve (AUC). Subsequently, the combined nomograms were constructed based on the radiographic and clinical features and verified.Results The AUC of the radiographic model in predicting prostate cancer in the training and validation sets was 0.946 (95% CI: 0.903-0.982) and 0.902 (95% CI: 0.862-0.958). AUC comparable with pooled models was 0.965 (95% CI: 0.904-0.989) and 0.924 (95% CI: 0.868-0.980), respectively.Conclusions Bp-MRI radiomics model has high diagnostic efficiency for prostate cancer (PCa). The combined nomograms that combine total prostate specific antigen (tPSA), free prostate specific antigen (fPSA)/tPSA (f/t), and radiographic features may provide an effective tool for risk prediction and individualized treatment in patients with prostate disease.
[关键词] 双参数磁共振成像;影像组学;前列腺癌;诊断效能;列线图;前列腺特异性抗原
[Keywords] biparameter magnetic resonance imaging;radiomics;prostate cancer;diagnostic efficacy;nomogram;prostate specific antigen

赵莹莹 1, 2   方陈 3   吴声连 1   徐伟 1   郑鹏翔 3   郑伟龙 1   陈志强 2*  

1 福建医科大学附属福清市医院影像科,福州 350000

2 宁夏医科大学总医院放射科,银川 750004

3 福建医科大学附属福清市医院泌尿外科,福州 350000

陈志强,E-mail:zhiqiang_chen99@163.com

作者利益冲突声明:全部作者均声明无利益冲突。


基金项目: 福建省卫生健康科技计划项目 2021QNA067 宁夏回族自治区重点研发计划项目 2019BEG03033 宁夏自然科学基金 2022AAC03472
收稿日期:2022-02-19
接受日期:2022-07-28
中图分类号:R445.2  R737.25 
文献标识码:A
DOI: 10.12015/issn.1674-8034.2022.08.008
本文引用格式:赵莹莹, 方陈, 吴声连, 等. 基于Bp-MRI影像组学预测前列腺病变良恶性的效能及风险评估[J]. 磁共振成像, 2022, 13(8): 43-47. DOI:10.12015/issn.1674-8034.2022.08.008.

       前列腺癌(prostate cancer, PCa)是一种“老化的肿瘤”,约六分之一的男性一生中会罹患[1, 2],并且发病较隐匿,早期发现和准确诊断可以提高患者生存率并降低治疗成本。一项大样本研究表明,在初次活检前使用MRI进行分诊可减少约四分之一的不必要活检,并避免临床上无意义病灶的过度诊断[3]。另一项研究表明,结合扩散加权成像(diffusion weighted imaging, DWI)和T2WI的双参数MRI(biparametric MRI, Bp-MRI)可以提高肿瘤诊断及其侵袭性识别的能力[4]。然而,传统的医学影像学有不可避免的局限性,通过肉眼获得的信息是有限的,只有少数特征被识别和描述,影响了诊断肿瘤良恶性和分期的准确性[4, 5]。此外,医学影像在很大程度上依赖于放射科医生的知识和经验,这导致了其高异质性和低重现性。

       影像组学可从医学影像中定量提取图像特征,使用算法或统计分析工具捕获肿瘤内不同区域之间的形态和功能的细微异质性,提供除主观定性分析之外的大量信息,比传统影像学更能准确地描述肿瘤[6]。目前,MRI影像组学已广泛应用于PCa的诊断与鉴别、病理分级、侵袭性评估以及疗效预测和预后分析的研究中[7, 8, 9, 10]。但大多数研究[7, 8, 9]只关注了病变的影像组学特征,未能充分利用临床指标的优势。因此,本文旨在探讨基于T2WI及DWI的Bp-MRI的影像组学结合临床危险因素的联合模型识别PCa的诊断性能,并建立一个个体化PCa预测模型。

1 材料与方法

1.1 临床资料

       本研究经福建医科大学附属福清市医院医学伦理委员会批准,免除受试者知情同意,批准文号:K(2021)14号。回顾性分析2018年6月至2022年1月在福建医科大学附属福清市医院行前列腺MRI检查的患者的临床信息[包括年龄、总前列腺特异性抗原(total prostate specific antigen, tPSA)、游离前列腺特异抗原(free prostate specific antigen, fPSA)/tPSA比值(f/t)等]及影像学资料。纳入标准:(1)前列腺MRI图像上有明显的病灶(病灶直径≥0.5 cm);(2)经病理学证实,并与MRI图像匹配在同一区域;(3)超声引导下前列腺穿刺或根治手术于MRI检查后一个月内进行。排除标准:(1)MRI检查前已经治疗或穿刺;(2)多种原发性癌症或既往有癌症病史;(3)序列不完整或图像质量不佳;(4)临床资料不完整。

1.2 图像获取、分析和数据采集

       所有患者均使用Siemens Spectra 3.0 T磁共振扫描仪,以体线圈为射频发射线圈,心脏相控阵线圈及脊柱线圈为接收线圈。T2WI序列扫描参数如下:TE 72 ms,TR 4000 ms,矩阵128×128,FOV 360 mm×360 mm,层厚4 mm,层间距0.5 mm;DWI序列扫描参数如下:TE 72 ms,TR 4100 ms,矩阵128×128,FOV 360 mm×360 mm,层厚4 mm层间距0.5 mm,b值为0、2000 s/mm2,激励次数0、10。

       由一名具有5年腹部MRI诊断经验的影像医师将入组患者的T2WI、DWI(b=2000 s/mm2)及ADC原始图像以医学数字成像和通信(digital imaging and communications in medicine, DICOM)格式导出并传入开源软件ITK-SNAP 3.8,根据三维感兴趣区法及详细的病理结果逐层勾画病灶区,见图1。然后使用Python 3.7.1的影像组学包(PyRadiomics 3.0)提取了3111个影像组学特征[每个序列各1037个特征,包括14个形状特征(original_shape),198个一阶直方图特征,264个灰度共生矩阵(gray level co-occurence matrix, GLCM)特征,154个灰度相关矩阵(gray level dependence matrix, GLDM)特征,176个灰度游程矩阵(graylevel run length matrix, GLRLM)特征,176个灰度区域大小矩阵(gray level size zone matrix, GLSZM)特征,55个邻域灰度差矩阵(neighbouring gray tone difference matrix, NGTDM)特征)]。由另1名具有8年腹部MRI诊断经验的影像医师随机抽取60例患者重复上述操作。以一致性相关系数(inter-class correlation coefficient, ICC)评价观察者间提取影像组学特征的一致性,ICC>0.75为一致性较好。

图1  男,75岁,病理诊断为前列腺癌。1A:病理图(HE ×100)示腺体融合、腺腔发育不全,并见筛状腺体,Gleason评分为4+4;1B:T2WI示外周带5~7点钟方向T2WI低信号结节;1C:ITK-SNAP软件勾画病灶区示意图;1D:扩散加权成像(DWI)图(b=2000 s/mm2),病灶呈明显高信号;1E:表观扩散系数(ADC)图,病灶呈明显低信号。
Fig. 1  Male, 75-year-old, pathologically diagnosed prostate cancer. 1A: The pathological image (HE ×100) showed glandular fusion, glandular agenesis, and sieve glands, Gleason score 4+4. 1B: T2WI image showed low signal nodules at 5-7 o'clock; 1C: Schematic diagram of the focal area delineated by ITK-SNAP; 1D: The diffusion weighted imaging (DWI) image (b=2000 s/mm2), the lesion showed a significantly high signal; 1E: The apparent diffusion coefficient (ADC) image, the lesion showed a significantly low signal.

1.3 特征筛选及模型建立

       首先对所有的影像组学特征进行数据标准化,并对训练集的数据进行一次上采样,避免样本不平衡。随后,将一致性较差(ICC<0.75)的特征过滤掉,再采用t检验或Wilcoxon秩和检验筛选出恶性组与良性组间差异有统计学意义的变量;然后,采用最小绝对收缩和选择算子(least absolute shrinkage and selection operator, LASSO)算法对非零系数的特征进行选择;接着,用Spearman相关分析计算特征之间的相关系数,剔除高度相关(r>0.7)的特征来减少冗余数据;再采用单因素logistic回归分析,选择具有显著性差异的特征(P<0.05);最后,采用多因素logistic回归分析,从剩余的特征中筛选出有价值的特征进行拟合,建立一个结合影像组学特征和临床特征的联合模型。

1.4 统计学方法

       所有统计分析均在R软件(v.3.5.1)和Python软件(v.3.7.1)进行。计数资料行χ2检验,对计量参数进行方差齐性检验,如符合正态分布,行两独立样本t检验,如不符合正态分布,则行非参数检验,P<0.05表示差异有统计学意义。通过Spearman相关分析来评估影像组学特征之间的相关性。采用logistic回归模型对筛选出的临床及组学特征进行建模分析。采用受试者工作特征(receiver operating characteristic, ROC)曲线及曲线下面积(area under the curve, AUC)和校准曲线来评估模型的性能。

2 结果

2.1 基本信息

       最终有161名患者纳入本研究,年龄50~87岁,平均年龄70.3岁。其中PCa患者59例[前列腺特异性抗原(prostate specific antigen, PSA)范围3.2~100.0 ng/mL,平均PSA水平为19.6 ng/mL],良性疾病患者102例(PSA范围0.6~43.8 ng/mL,平均PSA水平为5.7 ng/mL)。这些患者按照7∶3的比例随机、分层划分为训练组(n=113)和验证组(n=48)。

2.2 组学特征筛选及模型建立

       首先剔除掉572个一致性差(ICC<0.75)的组学特征,然后使用两独立样本t检验/Wilcoxon秩和检验筛选出两组间存在显著差异的1525个特征,用LASSO算法筛选出非零系数的9个特征并根据相应的特征系数计算出影像组学评分(Radscore)(具体参数信息见表1),再行Spearman相关分析发现有4个特征相关系数大于0.7,随机剔除两个特征:b2000.original_shape_Sphericity(b2000.OSS)、b2000.original_glszm_SmallAreaEmphasis(b2000.OGSAE),最终剩余7个特征:T2WI.original_shape_LeastAxisLength(T2WI.OSLA)、T2WI.original_shape_Sphericity(T2WI.OSS)、b2000.exponential_firstorder_Minimum(b2000.EFM)、b2000.wavelet.LLH_glszm_GrayLevelNonUniformity(b2000.WLGGN)、ADC.square_gldm_DependenceVariance(ADC.SGDV)、ADC.wavelet.LHL_glcm_Idn(ADC.WLI)、ADC.wavelet.LLL_firstorder_Median(WLFM),相关性矩阵见图2。影像组学模型在训练队列中AUC为0.946(95% CI:0.903~0.982),在验证队列中AUC为0.902(95% CI:0.862~0.958),ROC曲线见图3。采用单因素logistic回归分析对临床特征及影像组学特征进行选择和验证,选择P<0.05的变量,并计算各参数诊断PCa的诊断效能,见表1。然后,采用向后逐步logistic回归分析筛选出tPSA、f/t和Radscore作为PCa的显著预测因子并制作联合列线图(图4),在训练队列中AUC为0.965(95% CI:0.904~0.989),在验证队列中AUC为0.924(95% CI:0.868~0.980),高级校准曲线如图4所示。并利用联合列线图计算出每个病例对应患PCa的概率及最佳截断值,见表1

图2  7个影像组学特征的相关性矩阵图。OSS为original_shape_Sphericity特征;EFM为exponential_firstorder_Minimum特征;WLFM为wavelet.LLL_firstorder_Median特征;OSLA为original_shape_LeastAxisLength特征;WLGGN为wavelet.LLH_glszm_GrayLevelNonUniformity特征;SGDV为square_gldm_DependenceVariance特征;WLI为wavelet.LHL_glcm_Idn特征。
图3  影像组学模型在训练集和验证集诊断前列腺癌(PCa)的受试者工作特征(ROC)曲线。
图4  列线图、校准曲线及决策曲线。4A:基于结合临床危险因素总前列腺特异性抗原(tPSA)、fPSA/tPSA比值(fPSA为游离前列腺特异抗原)和影像组学评分Radscore的联合模型的列线图。4B、4C:训练集、验证集的校准曲线显示,联合模型在训练集及验证集中的校准度均较好,ROC曲线下面积(AUC)分别为0.965、0.924。y轴表示患者中PCa的实际患病率;x轴表示PCa的列线图预测概率;灰色对角实线表示由完美模型拟合的理想模型。4D:训练集中联合模型、组学模型和临床模型的决策曲线(DCA)显示,阈值取0~0.9时,联合模型在训练集数据中的临床净获益均大于临床模型。y轴表示净收益,x轴表示阈值概率。绿线表示联合模型,红线表示临床模型,蓝线表示组学模型,灰线表示所有患者均为PCa患者的假设。
Fig. 2  Correlation matrix plot of the 7 radiomics features. OSS: original_shape_Sphericity; EFM: exponential_firstorder_Minimum; WLFM: wavelet.LLL_firstorder_Median; OSLA: original_shape_LeastAxisLength ; WLGGN: wavelet.LLH_glszm_GrayLevelNonUniformity ; SGDV: square_gldm_DependenceVariance; WLI: wavelet.LHL_glcm_Idn.
Fig. 3  Receiver operating characteristic (ROC) curves of prostate cancer diagnosis in the training set and validation set.
Fig. 4  Nomogram, calibration curve and decision curve. 4A: Nomogram of the combined model based on combining clinical risk factors total prostate specific antigen (tPSA), free prostate specific antigen (fPSA)/tPSA (f/t) and Radscore. 4B, 4C: The calibration curves of the training set (4B) and the validation set (4C) show that the calibration of the joint model was good in both the training set and the validation set, with an area under the ROC curve of 0.965 and 0.924, respectively. The y-axis indicates the actual prevalence of prostate cancer (PCa) in patients; the x-axis represents the nomogram prediction probability of the PCa, and the gray diagonal solid line represents the ideal model fitted by a perfect model. 4D: Decision curve (DCA) of the combined model, omics model and clinical model in the training set show that when the threshold is 0 to 0.9, the net clinical benefit of the combined model is greater than the clinical model. The y-axis represents the net gain, and the x-axis represents the threshold probability. Green lines indicate the combined model, red lines indicate the clinical model, blue lines indicate the radiomics model, and grey lines indicate the assumption that all patients were PCa patients.
表1  临床特征及影像组学特征诊断前列腺癌的效能
Tab. 1  Efficacy of clinical characteristics and radiomics characteristics in the diagnosis of prostate cancer

3 讨论

       本研究中我们从Bp-MRI图像中提取并筛选出7个影像组学特征,结合临床危险因素(tPSA、f/t),构建了PCa预测模型并对其进行验证。结果显示,在训练队列和验证队列中均有良好的性能,AUC分别为0.965和0.924,均优于单独的临床模型或影像组学模型。这充分显示了影像组学和临床特征的互补性,影像组学在宏观水平上捕获肿瘤的表型,而肿瘤标志物(tPSA、f/t)可在微观水平上提供详细量化信息。

3.1 临床变量对PCa的诊断价值

       本研究结果显示,tPSA的表达水平及f/t与PCa呈显著相关,这与以往的研究一致[11, 12]。PSA是由前列腺上皮细胞分泌产生的丝氨酸蛋白酶和糖蛋白,并直接分泌到前列腺导管系统,正常前列腺管系统周围有一个血—上皮屏障,可阻止PSA直接从前列腺上皮进入血液,从而维持血液中PSA的低浓度[13]。PCa会破坏血—上皮屏障,导致血清PSA浓度明显增加,因此PSA是早期筛查和检测PCa的关键肿瘤标志物,并广泛应用于临床,但其缺乏特异性。研究表明,在PCa患者中绝大部分PSA为结合状态,f/t低于正常人或良性前列腺增生患者[14]。因此,临床会利用f/t来增加诊断信心,我们的结果显示f/t的最佳截断值为0.16,这与Thakur等[15]的报道一致。但仅基于这两个因素构建的临床模型的诊断性能较低。

3.2 影像组学是鉴别前列腺良恶性肿瘤疗效的有效工具

       影像组学是一门结合人工智能和医学影像的综合性学科,过去10年在医学图像分析方面取得了巨大突破,可提取高通量特征,将医学图像数据转化为定量和多维特征,反映人体组织在细胞和遗传水平的变化,并提供更详细的肿瘤生物学信息和肿瘤内部微环境的特征[16],从而直观、定量地描述病变的形态和病理学特征。研究表明,影像组学在多种肿瘤中的表现优于传统的医学影像[6,16]。特别是应用于前列腺MRI时,基于形状、纹理和强度特征的预测模型可以帮助PCa的诊断、分期及复发和侵袭性评估[9, 10]

       在所有的影像组学特征中小波特征(wavelet)占比最大,这与之前的研究结果一致[17, 18]。小波域的特征聚焦于不同的频率范围,将图像分割成多层次的细节成分,能够区分肿瘤微环境的差异,量化肉眼无法识别的肿瘤异质性[19, 20]。具体来说,高频特征反映了肿瘤的边缘和细节信息,而低频特征获得了肿瘤的轮廓信息,同时对噪声进行了过滤,因此具有较强的预测能力,是构建影像组学模型的重要组成部分[21, 22]。其次为形状特征,其旨在描述肿瘤的大小及三维形态等相关信息,本文中包括T2WI.OSLA、T2WI.OSS以及b2000.OSS的组学特征,表明瘤灶的形状、体积是重要的预测特征,对PCa的鉴别具有重要意义,与Bai等[18]的结果一致。此外,我们还发现了一个与生物学特征相对应的代表图像区域内体素强度分布的直方图特征(b2000.EFM),此特征在PCa和良性病变之间的差异可能归因于肿瘤内异质性(如血管生成、侵袭性等)[23]

       为了便于临床应用,我们建立了一个可视化、可量化和个体化工具联合列线图模型,其结合了影像及临床资料,更全面、准确地预测患PCa的概率,为临床医生进行癌症管理提供有效的治疗指导[20]。根据列线图预测概率,可将患者分为PCa高风险患者和低风险患者。对于那些低风险的患者,不仅可以避免不必要的医疗检查和过度治疗,以减轻随访费用,还能有效预防低级别PCa的进展。近年来,在临床医学中得到了广泛的应用,该模型表达清晰、简洁、易于理解并利于医患沟通。

3.3 本研究的局限性

       本研究有几个局限性。首先,没有对PCa患者外周带及中央腺体癌进行分类;其次,样本量较小;最后,这是一项单中心回顾性研究。未来需要通过多中心研究、扩大样本量并对癌灶位置进行分类,进行外部验证来评估结果的普遍性。

[1]
Bi WL, Hosny A, Schabath MB, et al. Artificial intelligence in cancer imaging: clinical challenges and applications[J]. CA Cancer J Clin, 2019, 69(2): 127-157. DOI: 10.3322/caac.21552.
[2]
Liu LZ, Tian ZQ, Zhang ZF, et al. Computer-aided detection of prostate cancer with MRI: technology and applications[J]. Acad Radiol, 2016, 23(8): 1024-1046. DOI: 10.1016/j.acra.2016.03.010.
[3]
Kapoor J, Lamb AD, Murphy DG. Re: diagnostic accuracy of multi-parametric MRI and TRUS biopsy in prostate cancer (PROMIS): a paired validating confirmatory study[J]. Eur Urol, 2017, 72(1): 151. DOI: 10.1016/j.eururo.2017.02.014.
[4]
Muehlematter UJ, Burger IA, Becker AS, et al. Diagnostic accuracy of multiparametric MRI versus 68Ga-PSMA-11 PET/MRI for extracapsular extension and seminal vesicle invasion in patients with prostate cancer[J]. Radiology, 2019, 293(2): 350-358. DOI: 10.1148/radiol.2019190687.
[5]
Cuocolo R, Cipullo MB, Stanzione A, et al. Machine learning applications in prostate cancer magnetic resonance imaging[J/OL]. Eur Radiol Exp, 2019 [2022-02-19]. https://www.ncbi.nlm.nih.gov/pmc/articles/PMC6686027. DOI: 10.1186/s41747-019-0109-2.
[6]
Li WF, Shakir TM, Zhao YM, et al. Radiomics analysis of [18F] FDG PET/CT for microvascular invasion and prognosis prediction in very-early and early-stage hepatocellular carcinoma[J]. Eur J Nucl Med Mol Imaging, 2021, 48(11): 3353-3354. DOI: 10.1007/s00259-021-05479-w.
[7]
刘艳, 徐青青, 陆洋, 等. 基于T2WI及ADC图影像组学模型对前列腺癌和前列腺增生的鉴别诊断价值[J]. 临床放射学杂志, 2021, 40(11): 2168-2173. DOI: 10.13437/j.cnki.jcr.2021.11.026.
Liu Y, Xu QQ, Lu Y, et al. The study on the value of radiomics model based on T2WI and ADC in distinguishing prostate cancer and benign prostatic hyperplasia[J]. J Clin Radiol, 2021, 40(11): 2168-2173. DOI: 10.13437/j.cnki.jcr.2021.11.026.
[8]
Makowski MR, Bressem KK, Franz L, et al. De novo radiomics approach using image augmentation and features from T1 mapping to predict gleason scores in prostate cancer[J]. Invest Radiol, 2021, 56(10): 661-668. DOI: 10.1097/RLI.0000000000000788.
[9]
Nketiah GA, Elschot M, Scheenen TW, et al. Utility of T2-weighted MRI texture analysis in assessment of peripheral zone prostate cancer aggressiveness: a single-arm, multicenter study[J/OL]. Sci Rep, 2021 [2022-02-19]. https://www.ncbi.nlm.nih.gov/pmc/articles/PMC7822867. DOI: 10.1038/s41598-021-81272-x.
[10]
Gugliandolo SG, Pepa M, Isaksson LJ, et al. MRI-based radiomics signature for localized prostate cancer: a new clinical tool for cancer aggressiveness prediction? Sub-study of prospective phase Ⅱ trial on ultra-hypofractionated radiotherapy (AIRC IG-13218)[J]. Eur Radiol, 2021, 31(2): 716-728. DOI: 10.1007/s00330-020-07105-z.
[11]
Sooriakumaran P, Nyberg T, Akre O, et al. Survival among men at high risk of disseminated prostate cancer receiving initial locally directed radical treatment or initial androgen deprivation therapy[J]. Eur Urol, 2017, 72(3): 345-351. DOI: 10.1016/j.eururo.2017.04.002.
[12]
Zhang W, Mao N, Wang Y, et al. A Radiomics nomogram for predicting bone metastasis in newly diagnosed prostate cancer patients[J/OL]. Eur J Radiol. 2020 [2022-02-19]. https://www.ejradiology.com/article/S0720-048X(20)30209-6/fulltext. DOI: 10.1016/j.ejrad.2020.109020.
[13]
Li ST, Shi M, Zhao JJ, et al. A highly sensitive capillary electrophoresis immunoassay strategy based on dual-labeled gold nanoparticles enhancing chemiluminescence for the detection of prostate-specific antigen[J]. Electrophoresis, 2017, 38(13/14): 1780-1787. DOI: 10.1002/elps.201600396.
[14]
Tamimi W, Dafterdar R, Mansi M, et al. Complexed and total PSA in patients with benign prostatic hyperplasia and prostate cancer[J]. Br J Biomed Sci, 2010, 67(4): 184-188. DOI: 10.1080/09674845.2010.11730317.
[15]
Thakur V, Singh PP, Talwar M, et al. Utility of free/total prostate specific antigen (f/t PSA) ratio in diagnosis of prostate carcinoma[J]. Dis Markers, 2004, 19(6): 287-292. DOI: 10.1155/2004/913870.
[16]
Jiang YM, Yuan QY, Lv WB, et al. Radiomic signature of 18F fluorodeoxyglucose PET/CT for prediction of gastric cancer survival and chemotherapeutic benefits[J]. Theranostics, 2018, 8(21): 5915-5928. DOI: 10.7150/thno.28018.
[17]
Xia W, Hu B, Li HQ, et al. Multiparametric-MRI-based radiomics model for differentiating primary central nervous system lymphoma from glioblastoma: development and cross-vendor validation[J]. J Magn Reson Imaging, 2021, 53(1): 242-250. DOI: 10.1002/jmri.27344.
[18]
Bai HL, Xia W, Ji XF, et al. Multiparametric magnetic resonance imaging-based peritumoral radiomics for preoperative prediction of the presence of extracapsular extension with prostate cancer[J]. J Magn Reson Imaging, 2021, 54(4): 1222-1230. DOI: 10.1002/jmri.27678.
[19]
Aerts HJ, Velazquez ER, Leijenaar RT, et al. Decoding tumour phenotype by noninvasive imaging using a quantitative radiomics approach[J/OL]. Nat Commun. 2014 [2022-02-19]. https://www.ncbi.nlm.nih.gov/pmc/articles/PMC4059926. DOI: 10.1038/ncomms5006.
[20]
Gu DS, Xie YS, Wei JW, et al. MRI‐based radiomics signature: a potential biomarker for identifying glypican 3‐positive hepatocellular carcinoma[J]. J Magn Reson Imaging, 2020, 52(6): 1679-1687. DOI: 10.1002/jmri.27199.
[21]
Meng XC, Xia W, Xie PY, et al. Preoperative radiomic signature based on multiparametric magnetic resonance imaging for noninvasive evaluation of biological characteristics in rectal cancer[J]. Eur Radiol, 2019, 29(6): 3200-3209. DOI: 10.1007/s00330-018-5763-x.
[22]
Coroller TP, Grossmann P, Hou Y, et al. CT-based radiomic signature predicts distant metastasis in lung adenocarcinoma[J]. Radiother Oncol, 2015, 114(3): 345-350. DOI: 10.1016/j.radonc.2015.02.015.
[23]
Magistri P, Leonard SY, Tang CM, et al. The glypican 3 hepatocellular carcinoma marker regulates human hepatic stellate cells via Hedgehog signaling[J]. J Surg Res, 2014, 187(2): 377-385. DOI: 10.1016/j.jss.2013.12.010.

上一篇 基于ADC诊断局部晚期直肠癌新辅助治疗后完全缓解准确性的Meta分析
下一篇 T2 mapping联合mDixon-Quant多参数成像定量评估子宫内膜癌微卫星不稳定状态的研究
  
诚聘英才 | 广告合作 | 免责声明 | 版权声明
联系电话:010-67113815
京ICP备19028836号-2