分享:
分享到微信朋友圈
X
综述
人工智能在脑胶质瘤MRI诊断中的研究进展
赵维维 孙静 诸静其

Cite this article as: Zhao WW, Sun J, Zhu JQ. Research progress of artificial intelligence in MRI diagnosis of glioma[J]. Chin J Magn Reson Imaging, 2021, 12(8): 88-90.引用本文:赵维维, 孙静, 诸静其. 人工智能在脑胶质瘤MRI诊断中的研究进展[J]. 磁共振成像, 2021, 12(8): 88-90. DOI:10.12015/issn.1674-8034.2021.08.019.


[摘要] 脑胶质瘤是最常见的颅内原发性肿瘤,多呈浸润性生长,手术难以彻底切除,远处转移和对放化疗不敏感者治愈率极低,复发率高。患者长期生存率仅为20%。核磁共振是脑胶质瘤的首选检查方法,基于MRI的多模态影像学技术在脑胶质瘤的诊断和鉴别诊断、治疗前评估、手术导航及治疗后随访方面起到了关键的作用。高年资放射科医师通过MRI能较准确地识别肿瘤,但对于低年资放射科医师及未接受过脑肿瘤MRI诊断系统训练的放射科医生,误诊和漏诊的发生率明显升高。另外,放射科医师面对大量的MRI影像时常会出现身心疲劳,降低诊断准确性,因此,如何将放射科医师从传统阅片中解放出来,成为一个关注的问题。利用人工智能模拟人类思维,高效进行数据挖掘整合,从而实现精确诊断、鉴别诊断的数字医学出现迅猛发展。从影像大数据中提取肉眼难以有效识别的图像信息,通过分析这些信息来诊断疾病和建立预测模型,已成为具有广阔应用前景的技术手段。笔者就人工智能在脑胶质瘤MRI诊断中的研究进展做一综述,以提高对人工智能技术在脑胶质瘤鉴别诊断中的认识。
[Abstract] Glioma is the most common primary intracranial tumor with invasive growth, which is difficult to be completely removed by surgery. The cure rate of distant metastasis and insensitive to radiotherapy and chemotherapy is very low, and the recurrence rate is high. The long-term survival rate of patients is only 20%. MRI is the preferred method for the examination of brain glioma, and MRI-based multimodal imaging technology plays a key role in the diagnosis and differential diagnosis of brain glioma, pre-treatment evaluation, surgical navigation and post-treatment follow-up. However, for less experienced radiologists and those not trained in brain tumor MRI diagnostic system, the incidence of misdiagnosis and the missed diagnosis was significantly increased. In addition, radiologists often experience physical and mental fatigue in the face of a large number of MRI images, which reduces diagnostic accuracy. Therefore, how to liberate radiologists from the traditional reading of MRI images has become a concern. Artificial intelligence is used to simulate human thinking and efficiently carry out data mining and integration, so as to realize accurate diagnosis and differential diagnosis of digital medicine has developed rapidly. Extracting image information which is difficult to be effectively recognized by human eyes from image big data, and analyzing this information to diagnose disease and establish prediction model has become a technology with broad application prospects. In this paper, the research progress of artificial intelligence in MRI diagnosis of brain glioma is reviewed in order to improve the understanding of artificial intelligence in differential diagnosis of brain glioma.
[关键词] 人工智能;脑胶质瘤;磁共振成像
[Keywords] artificial intelligence;glioma;magnetic resonance imaging

赵维维 1   孙静 1   诸静其 2*  

1 同济大学附属普陀人民医院放射科,上海 200060

2 同济大学附属第十人民医院放射科,上海 200072

诸静其,E-mail:melvine0305@sina.com

全体作者均声明无利益冲突。


收稿日期:2021-01-10
接受日期:2021-02-22
DOI: 10.12015/issn.1674-8034.2021.08.019
引用本文:赵维维, 孙静, 诸静其. 人工智能在脑胶质瘤MRI诊断中的研究进展[J]. 磁共振成像, 2021, 12(8): 88-90. DOI:10.12015/issn.1674-8034.2021.08.019.

       脑胶质瘤(glioma)是最常见的颅内原发性肿瘤,多呈浸润性生长,手术难以彻底切除,远处转移和对放化疗不敏感者治愈率极低,复发率高,患者长期生存率仅为20%。根据世界卫生组织最新的肿瘤分类分级标准[1],脑胶质瘤在病理上分为少突胶质细胞肿瘤、星形细胞肿瘤、室管膜细胞瘤、脉络丛肿瘤、胚胎性肿瘤等,诊断与鉴别诊断上存在困难。MRI是脑胶质瘤的首选检查方法,基于MRI的多模态影像学技术在脑胶质瘤的诊断和鉴别诊断、治疗前评估、手术导航及治疗后随访等方面起到了关键作用。高年资放射科医师通过MRI能较准确地识别肿瘤,但对于低年资放射科医师及未接受过脑肿瘤MRI诊断系统训练的放射科医生,误诊和漏诊的发生率明显升高。另外,放射科医师面对大量的MRI影像时常会出现身心疲劳,降低了诊断准确性,因此,如何将放射科医师从传统阅片中解放出来,成为一个关注的问题。

       近年来,人工智能模拟人类思维,高效进行数据挖掘整合,从而实现精确诊断、鉴别诊断的数字医学出现迅猛发展。从影像大数据中提取肉眼难以有效识别的图像信息,通过分析这些信息来建立诊断疾病的预测模型,已成为具有广阔应用前景的技术手段。目前,人工智能已逐渐应用于脑肿瘤的MRI诊断和鉴别诊断[2, 3]。来自欧盟的一个联合项目[4]提出将高质量的数据集(MRI图像、临床、分子和遗传学),用于训练和机器学习的验证,然后用于表型(诊断)、治疗分配(预测)和预后,未来有望建立放射医学与精准医疗的桥梁。本文就人工智能在脑胶质瘤MRI诊断中的研究进展做一综述,以提高对人工智能技术在脑胶质瘤鉴别诊断中的认识。

1 人工智能算法

       目前,最常用的人工智能算法主要包括以下4种:人工神经网络(artificial neural networks,ANN)是一种可以构建模型的模拟人脑神经元网络的抽象算法;支持向量机(support vector machine,SVM)是一类按监督学习方式对数据进行二元分类的广义线性分类器,通过算法改进可以实现图像识别、文本分类等,较为常用;深度卷积神经网络(deep convolutional neural networks,CNNs)是深度学习的代表算法,其仿造生物的视知觉机制构建,常用于计算机视觉、自然语言处理,效果稳定;随机森林(random forest,RF)最早由贝尔实验室提出,是利用多棵树对样本进行训练并预测的一种高准确度的分类器[5, 6, 7]

2 人工智能在脑胶质瘤MRI诊断中的应用

2.1 鉴别肿瘤良恶性

       胶质瘤的良恶性程度和分级密切相关,根据WHO分级,高级别胶质瘤的恶性程度更高,预后极差[1]。快速鉴别良恶性肿瘤,可以避免恶性肿瘤漏诊而延误治疗时机,避免良性肿瘤接受不必要的手术治疗。人工智能技术通过图像预处理和增强、脑颅骨剥离、肿瘤分割、特征提取、特征选择可帮助医生精确实现良恶性肿瘤的分级。

       文献报道,通过SVM模型及逻辑回归模型和贝叶斯网络将脑胶质瘤增强图像特征整合至分类诊断预测模型中,对胶质瘤的分级诊断,其ROC可达到95%[5]。除外T1加权成像(T1 weighted imaging,T1WI)和T2加权成像(T2 weighted imaging,T2WI)图像,正电子发射断层扫描/磁共振成像(positron emission tomography/MRI,PET/MRI)结合人工智能也可对高低级别胶质瘤和基因突变状态进行很好的预测[8]。Zhuge等[9]利用CNNs原理对常规MRI图像进学习,通过两种不同的方法,分别实现了在不需要手术活检的情况下进行肿瘤分级,为有效的治疗方案和生存预测发挥了关键作用,而学习时间仅需数小时。这种方法是完全自动化的,不需要人工指定兴趣区域和选择切片进行模型训练。Gates等[10]使用4个参数对胶质瘤分级,准确率达96%,较常规MRI诊断准确率显著提高。Ahammed等[11]使用CNNs和wndchrm工具分类,准确率也可达到92.8%。高级别胶质瘤本质上是浸润性的,在磁共振成像液体衰减反转恢复序列(fluid attenuated inversion recovery,FLAIR)图像中血管源性水肿和非对比增强肿瘤均呈高信号,因此很难区分二者。Sengupta等[12]纳入高级别胶质瘤(high grade glioma,HGG)患者及转移患者术前后的T1WI、T2WI、FLAIR图像及动态对比增强MRI,采用SVM算法,结果在鉴别非增强性肿瘤和血管源性水肿方面,错误分类误差仅有2.4%。

       组蛋白H3 K27M突变肿瘤的类型比弥漫性固有桥脑胶质瘤的临床和放射学特征更能预测生存率。Jung等[13]通过回顾性纳入具有组蛋白H3 K27M突变型和野生型的41例胶质瘤患者MRI图像,采用RF模型进行分类,其准确度并不高。但Su等[14]在100例含有组蛋白H3 K27M突变型试验队列中得到的最终模型的ROC达到了0.9,显示了不错的生存率预测效果。异柠檬酸脱氢酶(isocitrate dehydrogenase,IDH)突变也是影响胶质瘤预后的因素之一,机器学习在预测脑胶质瘤IDH突变方面具有良好的诊断价值[15],整合临床和影像学特征较常规影像敏感度和特异度更高。基于扩散和灌注加权MRI的放射组学特征也能有效提高对低级别胶质瘤(low grade glioma,LGG)中IDH突变和肿瘤侵袭性的预测,具有ADC特征的多参数模型较常规模型表现出更高的性能[16]。扩散张量成像(diffusion tensor imaging,DTI)结合随机森林算法也能显著提高低级别胶质瘤IDH突变的分型[17]。该方案是完全自动化的,区分Ⅱ级胶质瘤和转移瘤的准确率最高,但区分Ⅱ级胶质瘤和Ⅲ级胶质瘤的准确率相对较低。也有学者通过遗漏交叉验证评估二元SVM算法[18, 19]分类脑转移瘤、脑膜瘤、Ⅱ级胶质瘤、Ⅲ级胶质瘤和胶质母细胞瘤。鉴别神经胶质瘤转移、高级别(Ⅲ级和Ⅳ级)肿瘤和低级别(Ⅱ级)肿瘤的有效性均可达80%以上。

2.2 鉴别肿瘤来源

2.2.1 鉴别脑转移瘤

       胶质母细胞瘤和实质器官的单发脑转移瘤在MRI上的表现相似,尽管DTI、磁共振波谱学(magnetic resonance spectroscopy,MRS)等一系列MRI技术已可以鉴别[18, 19],但有时仍难以区分,且耗时耗力。自动化计算机分析工具的可用性比放射科医师更客观,通过常规MRI和灌注MRI相结合的计算机辅助分类方法,可用于鉴别诊断。

       Artzi等[20]的一项回顾性研究中,仅通过常规的T1加权图像,使用了包括SVM、k近邻、决策树和集成分类器等4种人工智能算法进行分类,平均准确率为85%,实现了高效鉴别胶质母细胞瘤与脑转移瘤。对鉴别乳腺、肺等其他脑转移瘤亚型也有一定效果。Zacharaki等[21]提出了一种利用灌注MRI计算出的图谱来鉴别脑转移瘤的分类方案。方法为从中央和边缘肿瘤、水肿和坏死区域提取形状特征、图像强度和纹理特征。结果发现这是一种非常有前途的方法,可以对脑肿瘤进行客观和定量的评估。

2.2.2 鉴别其他颅脑肿瘤

       尽管MRI影像技术发展全面,但胶质母细胞瘤和原发性中枢神经系统淋巴瘤(primary central nervous system lymphoma,PCNSL)之间的影像鉴别仍常令人困惑。使用支持向量机算法可以达到鉴别诊断的目的[22]。一项来自76例胶质瘤患者的回顾性研究中[23],筛选增强的T1加权图像中的67个纹理特征,选取最具识别性的4个特征作为输入变量训练支持向量机分类器,通过测试图像验证,准确率达到了75%。Kunimatsu等[24]对经病理证实为脑胶质瘤和PCNSL的患者资料进行回顾性分析,在T1加权图像上采集的一阶特征及用灰度共生矩阵、灰度运行长度矩阵、灰度大小区域矩阵和多重灰度大小区域矩阵计算得到的二阶特征,证实基于MR图像的纹理特征显著增强了胶质瘤和PCNSL之间的差异。基于人工智能的放射组学方法对儿童后颅窝室管膜瘤和毛细胞星形细胞瘤的MRI纹理特征进行鉴别,证实了放射组学在临床辅助诊断中的地位[25]。结合肿瘤实体成分的ADC直方图、MRI图像和患者年龄,使用不同的分类算法如贝叶斯网络、随机森林、神经网络、支持向量机和线性多项式可以实现多种脑肿瘤的鉴别如脑转移瘤、血管母细胞瘤、毛细胞星形细胞瘤、室管膜瘤和成神经管细胞瘤的鉴别率,准确率高达90%。对儿童脑肿瘤(成神经管细胞瘤、毛细胞星形细胞瘤和室管膜瘤)的MRI图像中提取的三维纹理特征进行神经网络分类算法计算,也可以准确而无创地提高诊断效率,总体分类准确率提高了19%[26, 27]

2.3 鉴别治疗后相关效应和脑胶质瘤复发

       脑胶质瘤术后常需放疗,少数患者会出现不同程度的放射性脑损伤,影像检查可有效辨别胶质肿瘤复发与放射性脑损伤,是诊断胶质瘤术后、放疗后肿瘤复发与放射性脑损伤的有效方式。

       人工智能在脑胶质瘤术后复发和治疗相关效应(treatment-related effects,TRE)的鉴别也扮演重要的角色。Gao等[28]结合T1WI和T2 FLAIR减影图,对56例放化疗后进展可疑的高级别胶质瘤患者进行研究。通过3种支持向量机分类方法的模拟学习,发现人工智能算法效果显著优于单独使用T1WI或FLAIR区分TRE和胶质瘤复发,这对于放疗后胶质瘤患者的治疗管理至关重要。Wang等[29]通过对160例经病理证实的胶质瘤患者资料进行分析,收集MRI的纹理特征,使用多变量逻辑回归建立了预测肿瘤复发的模型。该模型纳入了15个特征组合,建模组和验证组均可以达到效果,综合模型具有良好的鉴别力,可显著鉴别肿瘤复发和放射性坏死。深度卷积神经网络在鉴别复发和TRE方面也有不错的表现,Bacchi等[30]利用CNNs基于DWI+FLAIR序列组合的模型准确率可达到82%。Tang等[31]开发了一种基于多序列MRI引导的全自动深度学习方法,深度特征融合模型(deep feature fusion model,DFFM)是一种多序列MRI引导的CNN模型,它同时迭代学习CT图像和多序列MRI的深度特征,然后将这两种深度特征结合起来产生分类结果,该深度特征融合模型划分结果可靠。除此之外,结合PET/MRI的放射组学特征提取不仅可用于评估放疗后相关效应,也可用于脑胶质瘤术后放疗的定位。确定较大的转移瘤,预测放疗后局部反应,以及区分放疗损伤与局部脑转移复发,准确率高达80%~90%[32]

       综上所述,人工智能的方法多样,各有优势和不足,Kocak等[33]使用自适应增强、k近邻、贝叶斯、神经网络、随机森林、随机梯度下降和支持向量机等多种算法分类预测低级别胶质瘤的1p/19q共缺失状态,发现各种算法差异无统计学意义,准确率均可达到80%。对于哪种人工智能算法最优,目前国内外学者尚未达成共识,通过优化不同的机器学习算法可以使预测精度最大化。例如SVM效果好也最常用,但RF在变量选择方面具有优势,不仅考虑每个变量的影响,还可以测量特征的重要性。

       总之,人工智能可以大幅度缩短MRI诊断脑胶质瘤的时间并减少误诊和漏诊的发生率。随着算法和模型的进一步优化和改进,基于MRI的人工智能有望在脑胶质瘤的诊断、鉴别诊断及预后随访评估上成为放射科医师的重要辅助诊断手段。相信随着5G技术的到来,基于多中心权威MRI数据制定的用于诊断脑胶质瘤的人工智能计算机辅助诊断系统将会有更加广阔的应用前景。

1
Wen PY, Huse JT. 2016 World Health Organization classification of central nervous system tumors[J]. Continuum (Minneap Minn), 2017, 23(6): 1531-1547. DOI: 10.1212/CON.0000000000000536.
2
Thust SC, Heiland S, Falini A, et al. Glioma imaging in Europe: A survey of 220 centres and recommendations for best clinical practice[J]. Eur Radiol, 2018, 28(8): 3306-3317. DOI: 10.1007/s00330-018-5314-5.
3
Booth TC, Williams M, Luis A, et al. Machine learning and glioma imaging biomarkers[J]. Clin Radiol, 2020, 75(1): 20-32. DOI: 10.1016/j.crad.2019.07.001.
4
Marti-Bonmati L, Alberich-Bayarri A, Ladenstein R, et al. PRIMAGE project: predictive in silico multiscale analytics to support childhood cancer personalised evaluation empowered by imaging biomarkers[J]. Eur Radiol Exp, 2020, 4(1): 22. DOI: 10.1186/s41747-020-00150-9.
5
Shaver MM, Kohanteb PA, Chiou C, et al. Optimizing neuro-oncology imaging: A review of deep learning approaches for glioma imaging[J]. Cancers (Basel), 2019, 11(6): 829. DOI: 10.3390/cancers11060829.
6
Lotan E, Jain R, Razavian N, et al. State of the art: Machine learning applications in glioma imaging[J]. AJR Am J Roentgenol, 2019, 212(1): 26-37. DOI: 10.2214/AJR.18.20218.
7
Senders JT, Harary M, Stopa BM, et al. Information-based medicine in glioma patients: A clinical perspective[J]. Comput Math Methods Med, 2018, 2018: 8572058. DOI: 10.1155/2018/8572058.
8
Haubold J, Demircioglu A, Gratz M, et al. Non-invasive tumor decoding and phenotyping of cerebral gliomas utilizing multiparametric (18)F-FET PET-MRI and MR Fingerprinting[J]. Eur J Nucl Med Mol Imaging, 2020, 47(6): 1435-1445. DOI: 10.1007/s00259-019-04602-2.
9
Zhuge Y, Ning H, Mathen P, et al. Automated glioma grading on conventional MRI images using deep convolutional neural networks[J]. Med Phys, 2020, 47(7): 3044-3053. DOI: 10.1002/mp.14168.
10
Gates E, Lin JS, Weinberg JS, et al. Imaging-based algorithm for the local grading of glioma[J]. AJNR Am J Neuroradiol, 2020, 41(3): 400-407. DOI: 10.3174/ajnr.A6405.
11
Ahammed MK, Rajendran VR, Glioma PJK. Tumor grade identification using artificial intelligent techniques[J]. J Med Syst, 2019, 43(5): 113. DOI: 10.1007/s10916-019-1228-2.
12
Sengupta A, Agarwal S, Gupta PK, et al. On differentiation between vasogenic edema and non-enhancing tumor in high-grade glioma patients using a support vector machine classifier based upon pre and post-surgery MRI images[J]. Eur J Radiol, 2018, 106: 199-208. DOI: 10.1016/j.ejrad.2018.07.018.
13
Jung JS, Choi YS, Ahn SS, et al. Differentiation between spinal cord diffuse midline glioma with histone H3 K27M mutation and wild type: comparative magnetic resonance imaging[J]. Neuroradiology, 2019, 61(3): 313-322. DOI: 10.1007/s00234-019-02154-8.
14
Su X, Chen N, Sun H, et al. Automated machine learning based on radiomics features predicts H3 K27M mutation in midline gliomas of the brain[J]. Neuro Oncol, 2020, 22(3): 393-401. DOI: 10.1093/neuonc/noz184.
15
Zhao J, Huang Y, Song Y, et al. Diagnostic accuracy and potential covariates for machine learning to identify IDH mutations in glioma patients: evidence from a meta-analysis[J]. Eur Radiol, 2020, 30(8): 4664-4674. DOI: 10.1007/s00330-020-06717-9.
16
Kim M, Jung SY, Park JE, et al. Diffusion- and perfusion-weighted MRI radiomics model may predict isocitrate dehydrogenase (IDH) mutation and tumor aggressiveness in diffuse lower grade glioma[J]. Eur Radiol, 2020, 30(4): 2142-2151. DOI: 10.1007/s00330-019-06548-3.
17
Park CJ, Choi YS, Park YW, et al. Diffusion tensor imaging radiomics in lower-grade glioma: improving subtyping of isocitrate dehydrogenase mutation status[J]. Neuroradiology, 2020, 62(3): 319-326. DOI: 10.1007/s00234-019-02312-y.
18
Holly KS, Fitz-Gerald JS, Barker BJ, et al. Differentiation of high-grade glioma and intracranial metastasis using volumetric diffusion tensor imaging tractography[J]. World Neurosurg, 2018, 120: e131-e141. DOI: 10.1016/j.wneu.2018.07.230.
19
Luts J, Suykens JAK, Van Huffel S, et al. Differentiation between brain metastases and glioblastoma multiforme based on MRI, MRS and MRSI[C]. Proceedings of the IEEE Symposium on Computer-Based Medical Systems, August 3-4, 2009, Albuquerque, New Mexico, USA. DOI: 10.1109/CBMS.2009.5255249.
20
Artzi M, Bressler I, Ben BD. Differentiation between glioblastoma, brain metastasis and subtypes using radiomics analysis[J]. J Magn Reson Imaging, 2019, 50(2): 519-528. DOI: 10.1002/jmri.26643.
21
Zacharaki EI, Wang S, Chawla S, et al. Classification of brain tumor type and grade using MRI texture and shape in a machine learning scheme[J]. Magn Reson Med, 2009, 62(6): 1609-1618. DOI: 10.1002/mrm.22147.
22
Alcaide-Leon P, Dufort P, Geraldo AF, et al. Differentiation of enhancing glioma and primary central nervous system lymphoma by texture-based machine learning[J]. AJNR Am J Neuroradiol, 2017, 38(6): 1145-1150. DOI: 10.3174/ajnr.A5173.
23
Kunimatsu A, Kunimatsu N, Yasaka K, et al. Machine learning-based texture analysis of contrast-enhanced MR imaging to differentiate between glioblastoma and primary central nervous system lymphoma[J]. Magn Reson Med Sci, 2019, 18(1): 44-52. DOI: 10.2463/mrms.mp.2017-0178.
24
Kunimatsu A, Kunimatsu N, Kamiya K, et al. Comparison between glioblastoma and primary central nervous system lymphoma using MR image-based texture analysis[J]. Magn Reson Med Sci, 2018, 17(1): 50-57. DOI: 10.2463/mrms.mp.2017-0044.
25
Li M, Wang H, Shang Z, et al. Ependymoma and pilocytic astrocytoma: Differentiation using radiomics approach based on machine learning[J]. J Clin Neurosci, 2020, 78: 175-180. DOI: 10.1016/j.jocn.2020.04.080.
26
Payabvash S, Aboian M, Tihan T, et al. Machine learning decision tree models for differentiation of posterior fossa tumors using diffusion histogram analysis and structural MRI findings[J]. Front Oncol, 2020, 10: 71. DOI: 10.3389/fonc.2020.00071.
27
Peng L, Parekh V, Huang P, et al. Distinguishing true progression from radionecrosis after stereotactic radiation therapy for brain metastases with machine learning and radiomics[J]. Int J Radiat Oncol Biol Phys, 2018, 102(4): 1236-1243. DOI: 10.1016/j.ijrobp.2018.05.041.
28
Gao XY, Wang YD, Wu SM, et al. Differentiation of treatment-related effects from glioma recurrence using machine learning classifiers based upon pre-and post-contrast T1WI and T2 FLAIR subtraction features: A two-center study[J]. Cancer Manag Res, 2020, 12: 3191-3201. DOI: 10.2147/CMAR.S244262.
29
Wang K, Qiao Z, Zhao X, et al. Individualized discrimination of tumor recurrence from radiation necrosis in glioma patients using an integrated radiomics-based model[J]. Eur J Nucl Med Mol Imaging, 2020, 47(6): 1400-1411. DOI: 10.1007/s00259-019-04604-0.
30
Bacchi S, Zerner T, Dongas J, et al. Deep learning in the detection of high-grade glioma recurrence using multiple MRI sequences: A pilot study[J]. J Clin Neurosci, 2019, 70: 11-13. DOI: 10.1016/j.jocn.2019.10.003.
31
Tang F, Liang S, Zhong T, et al. Postoperative glioma segmentation in CT image using deep feature fusion model guided by multi-sequence MRIs[J]. Eur Radiol, 2020, 30(2): 823-832. DOI: 10.1007/s00330-019-06441-z.
32
Kocher M, Ruge MI, Galldiks N, et al. Applications of radiomics and machine learning for radiotherapy of malignant brain tumors[J]. Strahlenther Onkol, 2020, 196: 856-867. DOI: 10.1007/s00066-020-01626-8.
33
Kocak B, Durmaz ES, Ates E, et al. Radiogenomics of lower-grade gliomas: machine learning-based MRI texture analysis for predicting 1p/19q codeletion status[J]. Eur Radiol, 2020, 30(2): 877-886. DOI: 10.1007/s00330-019-06492-2.

上一篇 水通道蛋白4与缺血性脑卒中的相关性及影像学研究进展
下一篇 多模态MRI技术在脑泡型包虫病中的研究现状与进展
  
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