分享:
分享到微信朋友圈
X
临床研究
磁共振影像组学鉴别Ⅱ级孤立性纤维瘤/血管外皮细胞瘤与血管瘤型脑膜瘤的价值
付圣莉 任延德 李向荣 马驰 张华 葛亚琼

Cite this article as: Fu SL, Ren YD, Li XR, et al. The value of magnetic resonance imaging in differentiating grade Ⅱ solitary fibrous tumor/hemangiopericytoma from angiomatous meningioma[J]. Chin J Magn Reson Imaging, 2022, 13(1): 15-20.本文引用格式:付圣莉, 任延德, 李向荣, 等. 磁共振影像组学鉴别Ⅱ级孤立性纤维瘤/血管外皮细胞瘤与血管瘤型脑膜瘤的价值[J]. 磁共振成像, 2022, 13(1): 15-20. DOI:10.12015/issn.1674-8034.2022.01.004.


[摘要] 目的 探讨基于多参数MRI影像组学特征鉴别颅内Ⅱ级孤立性纤维瘤/血管外皮细胞瘤(solitary fibrous tumor/hemangiopericytoma,SFT/HPC)与血管瘤型脑膜瘤(angiomatous meningioma,AM)的价值。材料与方法 回顾性分析经手术或病理证实的两个中心共68例Ⅱ级SFT/HPC患者及41例AM患者的术前磁共振成像资料,按7∶3的比例将患者随机分为训练组(n=77)和验证组(n=32)。图像进行标准化后,利用3D slicer软件于T1WI、FLAIR及T1WI增强轴位图像勾画感兴趣区(region of interest,ROI)并进行特征提取;采用最小冗余最大相关(minimum redundancy maximum relevance,mRMR)和最小绝对收缩和选择算子(least absolute shrinkage and selection operator,LASSO)算法进行降维,筛选最具诊断价值的影像组学特征,构建二元Logistic回归模型,并绘制受试者操作特征(receiver operating characteristic,ROC)曲线,评估其鉴别诊断效能。结果 从T1WI、FLAIR和T1WI增强序列中各提取16、13、12个影像组学特征,从多参数MRI序列中提取9个影像组学特征进行模型建立,4种模型的ROC曲线下面积(area under the curve,AUC)分别为:T1WI模型为0.98 (敏感度100%,特异度92.86%);FLAIR模型为0.92 (73.47%,100%),T1WI增强模型为0.89 (79.59%,85.19%),多参数模型为0.99 (98.04%,96.15%),在验证组中的准确率分别为87.50%、75.00%、68.75%和90.63%。结论 多参数MRI影像组学特征对Ⅱ级SFT/HPC与AM的鉴别效能优于单一序列,其中单序列中T1WI影像组学特征鉴别效能最高。
[Abstract] Objective To investigate the value of radiomics features with multi-parameter MRI images in differential diagnosis between intracranial grade Ⅱ solitary fibrous tumor/hemangiopericytoma (SFT/HPC) and angiomatous meningioma (AM).Materials and Methods: A total of 68 patients with grade Ⅱ SFT/HPC and 41 patients with AM confirmed by surgery or pathology were retrospectively analyzed from the First Affiliated Hospital of Qingdao University and Guangxi Medical University, all of the patients were performed T1WI, FLAIR and contrasted TIWI scan. The patients were randomly divided into training set (n=77) and validation set (n=32) in a ratio of 7∶3. After a normalization approach applied on the image, the region of interest (ROI) along the tumor edge step by step based on the axial image with 3D slicer software were sketched, then the radiomics features were extracted in the ROI with 3D slicer software. Minimum redundancy maximum relevance (mRMR) and least absolute shrinkage and selection operator (LASSO) regression were applied to reduce the dimension, then the radiomics features with the most diagnostic value were selected to build a binary Logistic regression model. The receiver operating characteristic (ROC) curves were used to evaluate the diagnostic performance of the model.Results 16, 13 and 12 radiomics features were extracted from T1WI, FLAIR and contrasted T1WI scan, respectively; additional 9 radiomics features were extracted from the combined sequence for modeling. The ROC analyses on four models resulted in an area under the curve (AUC) of 0.98 (sensitivity 100%, specificity 92.86%) for T1WI model, 0.92 (73.47%, 100%) for FLAIR model, 0.89 (79.59%, 85.19%) for contrasted T1WI model, and 0.99 (98.04%, 96.15%) for the combined sequence model and were enough to correctly distinguish the two groups in 87.50%、75.00%、68.75% and 90.63% of cases in test set, respectively.Conclusions The differentiation efficiency of multi-parameter MRI images radiomics features between intracranial grade Ⅱ SFT/HPC and AM was better than single sequence. T1WI was the highest diagnosis efficacy sequence among single sequence.
[关键词] 磁共振成像;血管外皮细胞瘤;血管瘤型脑膜瘤;影像组学;鉴别效能
[Keywords] magnetic resonance imaging;hemangiopericytoma;angiomatous meningioma;radiomics;differentiation performance

付圣莉 1   任延德 1*   李向荣 2   马驰 1   张华 1   葛亚琼 3  

1 青岛大学附属医院放射科,青岛 266555

2 广西医科大学第一附属医院放射科,南宁 530000

3 通用电气药业(上海)有限公司,上海 210000

任延德,E-mail:8198458@163.com

全体作者均声明无利益冲突。


收稿日期:2021-08-22
接受日期:2021-12-29
中图分类号:R445.2  R730.262 
文献标识码:A
DOI: 10.12015/issn.1674-8034.2022.01.004
本文引用格式:付圣莉, 任延德, 李向荣, 等. 磁共振影像组学鉴别Ⅱ级孤立性纤维瘤/血管外皮细胞瘤与血管瘤型脑膜瘤的价值[J]. 磁共振成像, 2022, 13(1): 15-20. DOI:10.12015/issn.1674-8034.2022.01.004.

       孤立性纤维瘤/血管外皮细胞瘤(solitary fibrous tumor/hemangiopericytoma,SFT/HPC)在2016年的中枢神经系统分类中被分为三个级别,WHO Ⅱ级在过去被诊断为血管外皮细胞瘤(hemangiopericytoma,HPC)[1]。HPC是一类脑膜间叶组织来源肿瘤,仅占所有原发性中枢神经系统肿瘤的1%[2]。血管瘤型脑膜瘤(angiomatous meningioma,AM)是脑膜瘤的一种组织学亚型,其特征是高血管性肿瘤组织,WHO分级为Ⅰ级[3]。Ⅱ级SFT/HPC和AM具有相似的MRI表现,虽然已有研究显示MRI表现对SFT/HPC和AM的鉴别诊断具有一定的价值,如大小、信号强度、血管流空信号、脑膜尾征等[4, 5],但是往往会受到放射科医师的主观影响,因此常规MRI对两者的鉴别能力有限[6, 7],而相对于AM,Ⅱ级SFT/HPC具有较高的局部复发率和颅外转移[8],首次治疗最大程度的手术切除是影响患者预后及生存的最重要因素[9, 10]。因此,准确地术前鉴别诊断Ⅱ级SFT/HPC和AM在临床实践中具有重要意义。

       影像组学(radiomics)通过从不同模态的影像中提取高通量的影像特征[11],一定程度上实现了肿瘤的诊断与鉴别诊断、肿瘤的异质性与预后的评价,已经广泛应用于中枢神经系统病变的研究[12, 13, 14],但国内外应用影像组学研究颅内SFT/HPC与AM的鉴别诊断尚较少,且这些研究多来自单中心。本研究通过对两个中心的Ⅱ级SFT/HPC和AM回顾性分析,旨在探讨多参数MRI影像组学特征在Ⅱ级SFT/HPC和AM的鉴别诊断价值。

1 材料与方法

1.1 一般资料

       回顾性收集2015年6月至2020年9月期间青岛大学附属医院、广西医科大学第一附属医院患者病例共109例,本研究经青岛大学附属医院医学伦理委员会批准(批准文号:QYFYwzLL26524)、广西医科大学第一附属医院医学伦理委员会批准[批准文号:2021 (KY-E-201)],免除受试者知情同意。纳入标准:(1)病理诊断为Ⅱ级SFT/HPC或AM;(2)均接受术前MRI检查且无图像伪影;(3)初次诊断且未经任何治疗。然后通过统一标准的图像存档与通信系统(picture archiving and communication system,PACS)将患者术前MRI图像以DICOM格式导出。排除标准:(1)既往有脑外科手术或活检史;(2)既往有颅内疾病史,如蛛网膜下腔出血、脑梗死等。

1.2 MR图像采集

       采用西门子SKYRA 3.0 T及GE Signa HDX 3.0 T 磁共振仪进行检查,患者仰卧位,采用标准头线圈,进行轴位扫描。西门子SKYRA 3.0 T扫描参数:T1WI:TR 1800 ms,TE 8.5 ms;FLAIR:TR 9000 ms,TE 85 ms;T1WI增强:TR1800 ms,TE 8.5 ms;层厚5 mm,层间距1 mm,FOV 230 mm×230 mm。GE Signa HDX 3.0 T扫描参数:T1WI:TR 1800 ms,TE 24 ms;FLAIR:TR 8000 ms,TE 165 ms;T1WI增强:TR 2250 ms,TE 24 ms;层厚5 mm,层间距1 mm,FOV 230 mm×230 mm。经肘静脉应用高压注射器注射0.2 mL/kg钆喷酸葡胺(拜耳先灵,德国)后行增强扫描。

1.3 图像分析

1.3.1 图像分割

       将T1WI增强图像以DICOM格式导入到3D slicer (Version:4.8.1,http://www.slicer.org/)软件,由1名具有7年神经系统疾病诊断经验的放射科医师逐层勾画病灶边缘,确定病灶边界,合成三维感兴趣区(region of interest,ROI);每一次勾画均由另1名具有25年神经系统疾病诊断经验的放射科医师进行核对。ROI包括整个肿瘤,不包含肿瘤周围水肿区、大血管、静脉窦和增强的脑膜等。选取肿瘤强化明显的时相,参考T1WI增强图像对FLAIR、T1WI图像逐层勾画病灶边缘,以确保三个序列的ROI范围大致一致。

1.3.2 图像标准化

       由于两个中心的MR扫描仪和重建参数不同,在提取特征之前,对MRI图像进行标准化处理,包括重采样、降噪及小波变换,重采样到1 mm×1 mm×1 mm的体素大小,并进行高斯滤波,sigma值分别为0.5、1.0、1.5[15]

1.3.3 特征提取

       应用3D slicer软件对109例患者治疗前MRI图像的特征进行提取,共提取1166个定量参数,包括形态特征、一阶特征、灰度共生矩阵(gray level co-occurrence matrix,GLCM)特征、灰度区域大小矩阵(gray level size zone matrix,GLSZM)特征、灰度行程长度矩阵(gray level run length matrix,GLRLM)特征、邻域灰度差矩阵(neighboring gray tone difference matrix,NGTDM)特征和灰度相关矩阵(gray level dependence matrix,GLDM)特征。

1.4 影像组学模型建立与验证

       对于特征筛选,采用最小冗余最大相关(minimum redundancy maximum relevance,mRMR)和最小绝对收缩和选择算子(least absolute shrinkage and selection operator,LASSO)算法进行降维,并采用Logistic回归模型筛选最优的影像组学特征。将选择的具有非零系数的特征各自的系数加权线性组合,获得每例患者的影像组学分数(Rad-分数)[16]。采用受试者操作特征(receiver operating characteristic,ROC)曲线评估模型鉴别Ⅱ级SFT/HPC和AM的效能,计算曲线下面积(area under the curve,AUC)、准确率、敏感度和特异度。

1.5 统计学分析

       采用R软件(Version:3.5.1,https://www.Rproject.org)和IBM SPSS Statistics 22.0进行统计学分析。符合正态分布的计量资料用x¯±s表示,两组间比较采用独立样本t检验。非正态分布资料采用中位数(上、下四分位数)表示,两组间比较采用Mann-Whitney秩和检验。计数资料以频数表示,组间计数资料的比较采用χ²检验或Fisher确切概率法。本研究Ⅱ级SFT/HPC与AM的发病年龄采用两独立样本t检验,性别采用Fisher确切概率法,以P<0.05为差异有统计学意义。

2 结果

2.1 一般资料

       本研究Ⅱ级SFT/HPC患者共68例、AM患者共41例,按7∶3的比例将患者随机分为训练组(n=77)和验证组(n=32)。其中青岛大学附属医院Ⅱ级SFT/HPC患者53例,AM患者29例;广西医科大学第一附属医院Ⅱ级SFT/HPC患者15例,AM患者12例。本研究两组患者的年龄差异有统计学意义(P<0.001),性别差异无统计学意义(P=1.000,表1)。

表1  Ⅱ级SFT/HPC和AM患者的临床资料比较
Tab. 1  Comparison of clinical data of patients with grade Ⅱ SFT/HPC and AM

2.2 影像组学特征

       每例患者的每个序列均得到1166个影像组学特征,经过mRMR和LASSO对高维数据进行降维,最终T1WI选择了16个系数非零的影像组学特征,FLAIR选择了13个特征,T1WI增强选择了12个特征,然后结合三个序列,选择了9个最具诊断价值的影像组学特征(图1表2)。

图1  多参数MRI图像。A:二项式偏差随参数λ变化曲线图;B:影像组学特征系数随λ变化的曲线图。注:使用LASSO回归对影像组学特征进行筛选,利用交叉验证选择最优模型参数λ。纵轴为二项式偏差,横轴表示log (λ)值。上方数字表示筛选出特征的数量。垂直虚线表示实现最小二项式偏差的最佳λ取值所对应的log (λ)值(A);垂直虚线对应使用交叉验证后所筛选出的非零特征(B)。
Fig. 1  Multi-parameter MRI images. The curve of binomial deviance changing with parameter λ (A) and the curve of the radiomics features coefficients as λ changes (B) (Multi-parameter MRI images). LASSO regression was used to screen the radiomics features, and cross validation was used to select the optimal model parameter λ. The y-axis shows the binomial deviance, the x-axis shows the log (λ). The top number indicates the number of screened features. The vertical dashed line represents the value of log (λ) corresponding to the best value of λ for the smallest binomial deviance (A); vertical dashed lines correspond to non-zero features screened out using cross validation (B).
表2  不同参数MRI序列系数非零的影像组学特征
Tab. 2  Radiomics features of non-zero coefficients of MRI sequence with different parameters

2.3 影像组学特征鉴别Ⅱ级SFT/HPC与AM的效能

2.3.1 T1WI影像组学模型鉴别效能

       T1WI的训练组中影像组学特征模型鉴别Ⅱ级SFT/HPC与AM的AUC为0.98,95%可信区间(confidence interval,CI):0.96~1.00。验证组中AUC为0.94,95% CI:0.86~1.00 (图2)。

图2  基于T1WI图像影像组学模型鉴别诊断Ⅱ级SFT/HPC与AM的ROC曲线。
图3  基于FLAIR图像影像组学模型鉴别诊断Ⅱ级SFT/HPC与AM的ROC曲线。注:SFT/HPC:孤立性纤维瘤/血管外皮细胞瘤;AM:血管瘤型脑膜瘤。
图4  基于T1WI增强影像组学模型鉴别诊断Ⅱ级SFT/HPC与AM的ROC曲线。注:SFT/HPC:孤立性纤维瘤/血管外皮细胞瘤;AM:血管瘤型脑膜瘤。
图5  基于多参数MRI图像影像组学模型鉴别诊断Ⅱ级SFT/HPC与AM的ROC曲线。注:SFT/HPC:孤立性纤维瘤/血管外皮细胞瘤;AM:血管瘤型脑膜瘤。
图6  多参数MRI图像影像组学模型的决策曲线分析(DCA),纵轴为标准净收益,横轴为阈值概率。
Fig. 2  The ROC curves of grade Ⅱ SFT/HPC and AM were differentiated based on the radiomics model of T1WI images.
Fig. 3  The ROC curves of grade Ⅱ SFT/HPC and AM were differentiated based on the radiomics model of FLAIR images.
Fig. 4  The ROC curves of grade Ⅱ SFT/HPC and AM were differentiated based on the radiomics model of contrast-enhanced T1WI images.
Fig. 5  The ROC curves of grade Ⅱ SFT/HPC and AM were differentiated based on the radiomics model of multi-parameter MRI images.
Fig. 6  Decision Curve Analysis (DCA) of multi-parameter MRI images radiomics model. The y-axis shows the standardized net benefit, the x-axis shows the threshold probability.

2.3.2 FLAIR影像组学模型鉴别效能

       FLAIR的训练组中影像组学特征模型鉴别Ⅱ级SFT/HPC与AM的AUC为0.92,95% CI:0.87~0.98。验证组中AUC为0.91,95% CI:0.81~1.00 (图3)。

2.3.3 T1WI增强影像组学模型鉴别效能

       T1WI增强的训练组中影像组学特征模型鉴别Ⅱ级SFT/HPC与AM的AUC为0.89,95% CI:0.81~0.97。验证组中AUC为0.86,95% CI:0.73~0.99 (图4)。

2.3.4 多参数MRI序列影像组学模型鉴别效能

       T1WI、FLAIR及T1WI增强联合的训练组中影像组学特征模型鉴别Ⅱ级SFT/HPC与AM的AUC为0.99,95% CI:0.97~1.00。验证组中AUC为0.98,95% CI:0.95~1.00 (图5)。

       各个序列的训练组和验证组的敏感度、特异度、准确率、AUC值见表3所示。

表3  模型鉴别诊断Ⅱ级SFT/HPC与AM的效能
Tab. 3  Effectiveness of the model in differential diagnosis of grade Ⅱ SFT/HPC and AM

2.4 决策曲线分析(decision curve analysis, DCA)评估模型的临床应用价值

       本研究应用DCA来评估模型的临床应用价值,多参数联合序列模型显示在0.05~1的阈值概率下,模型可为患者带来临床净收益,“None”表示不给予相关预测及干预时患者无法获得临床收益;“All”表示患者接受预测或干预时获得的临床收益(图6)。

3 讨论

       本研究基于T1WI、FLAIR及TIWI增强序列获得不同的影像组学特征,建立影像组学模型鉴别Ⅱ级SFT/HPC和AM,旨在提高两者术前诊断准确率,结果显示在训练组及验证组中均得到较好的鉴别效能,在DCA中表现出较高的临床实用性。

3.1 相关研究比较

       国内外已有研究探究影像组学在颅内SFT/HPC与AM鉴别诊断中的价值。Wei等[5]比较临床数据模型、影像组学模型及临床数据结合影像组学模型对SFT/HPC及AM的鉴别效能,结果显示临床数据结合影像组学模型优于单一模型。Dong等[17]建立多参数MRI影像组学模型鉴别诊断SFT/HPC与AM,在训练组及验证组中表现出良好的鉴别效能。Li等[18]基于MRI序列建立的影像组学模型比基于临床特征构建的模型更好地鉴别SFT/HPC与AM,并且高于放射科医师的鉴别效能。上述研究均来自单中心,还需多中心数据来验证模型的泛化性。本研究收集了两个中心的数据,探讨多参数MRI影像组学特征鉴别诊断Ⅱ级SFT/HPC和AM的价值。

3.2 两组肿瘤不同序列影像组学特征模型对比分析

       本研究结果表明多参数MRI影像组学特征模型能较好地鉴别Ⅱ级SFT/HPC和AM,训练组及验证组AUC分别为0.99、0.98。多参数MRI序列结合了不同序列的特点,因此具有更高的鉴别诊断效能[19, 20]。Dong等[17]结合T1WI、T2WI、T1WI增强序列得到多参数影像组学模型,较准确地鉴别SFT/HPC和AM (训练组和验证组的准确率为0.887、0.831),表明多参数影像组学模型优于单一序列模型,与我们的结果一致。

       单个序列结果显示基于T1WI影像组学特征诊断效能优于FLAIR及T1WI增强,训练组及验证组AUC分别为0.98、0.94,且在多参数MRI影像组学特征中,T1WI特征所占比例最大,该结果可能由于两种肿瘤在T1WI上异质性更明显,如分叶状、坏死囊变等,与文献报道Ⅱ级SFT/HPC多发生坏死囊变,且多呈分叶状相符[21, 22]。T1WI增强图像能较好地反映肿瘤内部异质性及结构(如肿瘤血管生成)[23],本研究中训练组及验证组AUC分别为0.89、0.86,在三个序列中最低,Dong等[17]研究中T1WI增强模型的AUC最低,与我们的结果一致,可能由于两种肿瘤内部新生血管不丰富,结构异质性不明显。Li等[18]从FLAIR、DWI及T1WI增强三个序列各提取498个影像组学特征,构建支持向量机(support vector machine,SVM)分类器鉴别SFT/HPC和AM,结果显示T1WI增强AUC最高,可能由于SFT/HPC多表现为蛇形血管,AM为根状血管,结构异质性较大。

3.3 两组肿瘤影像组学特征模型临床应用价值

       本研究还应用DCA来评估模型的临床应用价值,DCA是一种通过净收益高低评估模型可靠性的方法,由Vickers和Elkin在2006年提出[24]。本研究中多参数MRI影像组学模型的决策曲线分析表明,在较大范围的阈值概率下,该模型具有较高的标准净收益,表明该模型具有较高的临床应用价值。Wei等[5]通过将临床影像学特征与影像组学特征相结合鉴别SFT/HPC和AM,并应用DCA来评估临床应用价值,结果显示,在训练组中,模型的净改善率为0.21%,在验证组中,模型的净改善率为0.19%,表明模型具有较好的临床实用性。

3.4 局限性

       本研究还存在以下不足:首先,这是一个回顾性的研究,可能存在选择偏倚;其次,因本研究中的病例为罕见病例,故收集的数据量较少,还需要更多数据量来进行验证;最后,由于部分患者缺少T2WI,故本研究仅纳入T1WI、FLAIR及T1WI增强序列数据进行研究,在后期研究中,我们会补充收集T2WI序列对研究结果进行验证。

       综上所述,相比传统的影像学方法,基于影像组学的方法能更加全面、客观、准确地鉴别Ⅱ级SFT/HPC和AM,帮助临床医师制订最佳的治疗方案,帮助患者改善预后。

[1]
Louis DN, Perry A, Reifenberger G, et al. The 2016 World Health Organization Classification of Tumors of the Central Nervous System: a summary[J]. Acta Neuropathol, 2016, 131(6): 803-820. DOI: 10.1007/s00401-016-1545-1.
[2]
Shankar JJS, Hodgson L, Sinha N.Diffusion weighted imaging may help differentiate intracranial hemangiopericytoma from meningioma[J]. J Neuroradiol, 2019, 46(4): 263-267. DOI: 10.1016/j.neurad.2018.11.002.
[3]
Hua L, Luan S, Li H, et al. Angiomatous Meningiomas Have a Very Benign Outcome Despite Frequent Peritumoral Edema at Onset[J]. World Neurosurg, 2017, 108: 465-473. DOI: 10.1016/j.wneu.2017.08.096.
[4]
He L, Li B, Song X, et al. Signal value difference between white matter and tumor parenchyma in T1- and T2- weighted images may help differentiating solitary fibrous tumor/hemangiopericytoma and angiomatous meningioma[J]. Clin Neurol Neurosurg, 2020, 198: 106221. DOI: 10.1016/j.clineuro.2020.106221.
[5]
Wei J, Li L, Han Y, et al. Accurate Preoperative Distinction of Intracranial Hemangiopericytoma From Meningioma Using a Multihabitat and Multisequence-Based Radiomics Diagnostic Technique[J]. Front Oncol, 2020, 10: 534. DOI: 10.3389/fonc.2020.00534.
[6]
El-Ali AM, Agarwal V, Thomas A, et al. Clinical metric for differentiating intracranial hemangiopericytomas from meningiomas using diffusion weighted MRI[J]. Clin Imaging, 2019, 54: 1-5. DOI: 10.1016/j.clinimag.2018.10.018.
[7]
梁小红, 周青, 赵志勇, 等. DWI联合最小ADC值对鉴别诊断颅内孤立性纤维性肿瘤/血管外皮细胞瘤与脑膜瘤的价值[J]. 磁共振成像, 2019, 10(1): 8-13. DOI: 10.12015/issn.1674-8034.2019.01.002.
Liang XH, Zhou Q, Zhao ZY, et al. The value of DWI combined with minimum ADC value in differential diagnosis of intracranial solitary fibrous tumor/hemangiopericytoma and meningioma. Chin J Magn Reson Imaging, 2019, 10(1): 8-13. DOI: 10.12015/issn.1674-8034.2019.01.002.
[8]
Wei G, Kang X, Liu X, et al. Intracranial meningeal hemangiopericytoma: Recurrences at the initial and distant intracranial sites and extraneural metastases to multiple organs[J]. Mol Clin Oncol, 2015, 3(4): 770-774. DOI: 10.3892/mco.2015.537.
[9]
Yip CM, Hsu SS, Liao WC, et al. Intracranial solitary fibrous tumor/hemangiopericytoma - A case series[J]. Surg Neurol Int, 2020,11: 414. DOI: 10.25259/SNI_490_2020.
[10]
Chen LF, Yang Y, Yu XG, et al. Multimodal treatment and management strategies for intracranial hemangiopericytoma[J]. J Clin Neurosci, 2015, 22(4): 718-25. DOI: 10.1016/j.jocn.2014.11.011.
[11]
Park YW, Choi YS, Ahn SS, et al. Radiomics MRI Phenotyping with Machine Learning to Predict the Grade of Lower-Grade Gliomas: A Study Focused on Nonenhancing Tumors[J]. Korean J Radiol, 2019, 20(9): 1381-1389. DOI: 10.3348/kjr.2018.0814.
[12]
Gore S, Chougule T, Jagtap J, et al. A Review of Radiomics and Deep Predictive Modeling in Glioma Characterization[J]. Acad Radiol, 2021, 28(11): 1599-1621. DOI: 10.1016/j.acra.2020.06.016.
[13]
Zhu Y, Man C, Gong L, et al. A deep learning radiomics model for preoperative grading in meningioma[J]. Eur J Radiol, 2019, 116: 128-134. DOI: 10.1016/j.ejrad.2019.04.022.
[14]
Bae S, An C, Ahn SS, et al. Robust performance of deep learning for distinguishing glioblastoma from single brain metastasis using radiomic features: model development and validation[J]. Sci Rep, 2020, 10(1): 12110. DOI: 10.1038/s41598-020-68980-6.
[15]
Wang J, Liu X, Hu B, et al. Development and validation of an MRI-based radiomic nomogram to distinguish between good and poor responders in patients with locally advanced rectal cancer undergoing neoadjuvant chemoradiotherapy[J]. Abdom Radiol (NY), 2021, 46(5): 1805-1815. DOI: 10.1007/s00261-020-02846-3.
[16]
陈海妹, 刘金, 程梓轩, 等. 基于多中心T1WI影像组学列线图治疗前预测骨肉瘤一年内复发的价值[J]. 中华放射学杂志, 2020, 54(9): 874-881. DOI: 10.3760/cma.j.cn112149-20200512-00675.
Chen HM, Liu J, Cheng ZX, et al. Value of radiomics nomogram based on T1WI for pretreatment prediction of relapse within 1 year in osteosarcoma: a multicenter study[J]. Chin J Radiol, 2020, 54(9): 874-881. DOI: 10.3760/cma.j.cn112149-20200512-00675.
[17]
Dong J, Yu M, Miao Y, et al. Differential Diagnosis of Solitary Fibrous Tumor/Hemangiopericytoma and Angiomatous Meningioma Using Three-Dimensional Magnetic Resonance Imaging Texture Feature Model[J]. Biomed Res Int, 2020, 2020: 5042356. DOI: 10.1155/2020/5042356.
[18]
Li X, Lu Y, Xiong J, et al. Presurgical differentiation between malignant haemangiopericytoma and angiomatous meningioma by a radiomics approach based on texture analysis[J]. J Neuroradiol, 2019, 46(5): 281-287. DOI: 10.1016/j.neurad.2019.05.013.
[19]
Xia W, Hu B, Li HQ, et al. Deep Learning for Automatic Differential Diagnosis of Primary Central Nervous System Lymphoma and Glioblastoma: Multi-Parametric Magnetic Resonance Imaging Based Convolutional Neural Network Model[J]. J Magn Reson Imaging, 2021, 54(3): 880-887. DOI: 10.1002/jmri.27592.
[20]
Zhao YJ, Lu YP, Li XX, et al. The Evaluation of Radiomic Models in Distinguishing Pilocytic Astrocytoma From Cystic Oligodendroglioma With Multiparametric MRI[J]. J Comput Assist Tomogr, 2020, 44(6): 969-976. DOI: 10.1097/RCT.0000000000001088.
[21]
Chen C, Ren CP. Value of Apparent Diffusion Coefficient (ADC) and Dynamic Contrast-Enhanced Magnetic Resonance Imaging (DCE-MRI) in Differentially Diagnosing Angiomatous Meningiomas and Solitary Fibrous Tumors/Hemangiopericytomas[J]. Med Sci Monit, 2019, 25: 5992-5996. DOI: 10.12659/MSM.915308.
[22]
He L, Li BH, Song XY, et al. Signal value difference between white matter and tumor parenchyma in T1- and T2- weighted images may help differentiating solitary fibrous tumor/ hemangiopericytoma and angiomatous meningioma[J]. Clin Neurol Neurosurg, 2020, 198: 106221. DOI: 10.1016/j.clineuro.2020.106221.
[23]
Kandemirli SG, Chopra S, Priya S, et al. Presurgical detection of brain invasion status in meningiomas based on first-order histogram based texture analysis of contrast enhanced imaging[J]. Clin Neurol Neurosurg, 2020, 198: 106205. DOI: 10.1016/j.clineuro.2020.106205.
[24]
Vickers AJ, Van Calster B, Steyerberg EW. A simple, step-by-step guide to interpreting decision curve analysis[J]. Diagn Progn Res, 2019, 3: 18. DOI: 10.1186/s41512-019-0064-7.

上一篇 三维动脉自旋标记灌注成像显示学龄前孤独症儿童部分脑区血流下降
下一篇 T1-mapping 技术对肿瘤坏死因子-α 拮抗剂治疗中轴型脊柱关节病疗效监测的初步研究
  
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