【作者】 刘振；

【导师】 刘献志；

【作者基本信息】 郑州大学 ， 神经外科学， 2022， 硕士

【摘要】 背景脑胶质瘤是常见的颅内占位性病变之一,相较于其他颅内占位性病变,如脑膜瘤、听神经瘤及室管膜瘤等,具有较高的侵袭性及破坏性,不仅发病率较高、治疗最为复杂,而且难以治愈。其中,以WHO Ⅳ级胶质母细胞瘤(GBM)最为恶性,仅有15个月左右的中位生存期。此外,原发性脑胶质瘤中的肿瘤分子遗传学标志物与肿瘤患者的总生存时间、治疗反馈及预后状态等密切相关。例如,与异柠檬酸脱氢酶(IDH)野生型胶质瘤患者相比,IDH突变型的病人病情进展相对迟缓,拥有更长的生存期;而IDH野生型的胶质瘤患者病情进展迅速且手术切除后极易复发,预后状态更差。O6-甲基鸟嘌呤-DNA甲基转移酶(MGMT)启动子超甲基化与烷化剂化疗敏感性高度相关。端粒酶逆转录酶(TERT)启动子突变也提示原发性胶质母细胞瘤患者的预后更差。2016年世界卫生组织(WHO)中枢神经系统(CNS)肿瘤分类将IDH和1p/19q纳入胶质瘤分类特征内,根据IDH和1p/19q不同状态将成人型弥漫性胶质瘤分为三种分子亚型:IDH野生型(IDHwt),IDH突变和1p/19q非共缺失型(IDHmut-noncodel),IDH突变和1p/19q共缺失型(IDHmut-codel)。同时,IDH和1p/19q被认为是预测成人型弥漫性胶质瘤肿瘤分类、判断胶质瘤危险分层和选择手术后治疗方案的最重要分子遗传标记物之一。然而,胶质瘤组织内各类型遗传分子标志物信息的明确多需要组织活检或手术切除后组织病理诊断等有创检查。因此,寻找一种无创的、术前的、高准确性的预测胶质瘤遗传分子标志物的方法显得尤为重要。近年来,依托于磁共振成像常规序列(T1WI,T2WI,CE-T1WI,FLAIR和ADC等)的影像组学研究对脑肿瘤的分级、分型、预后及生存进行了深入的探讨,并取得了显著的成果。作为磁共振成像中的功能成像序列,动态磁敏感灌注加权成像(DSC-PWI),在描述胶质瘤内血管网络及其血供特性上具有明显的优势性,能够清晰地反映出胶质瘤的血管密集程度及新生血管增殖能力。因此,我们的目标是利用动态磁敏感灌注加权成像中提取的影像组学特征建立模型来预测成人型弥漫性胶质瘤的三种分子亚型,从而对肿瘤进行早期干预,以提高患者的生存及预后。方法此回顾性研究收集了 2013年1月到2020年1月之间,在郑州大学第一附属医院手术切除后经组织病理确认的中枢神经系统胶质瘤患者的临床及影像资料信息,共1728例。经过入组筛选后共纳入278名成人型弥漫性胶质瘤患者。从动态磁敏感灌注加权成像中获得脑血容量(CBV)、脑血流量(CBF)、平均传输时间(MTT)和到达峰值时间(TTP)参数图,校准后从图像中提取影像组学特征。进行特征去冗余和特征筛选后,依托随机森林算法,利用最高相关性特征集构建影像组学模型,并将纳入患者按照13:7的比例随机分为训练集(n=175)和验证集(n=103)。最终,影像组学模型的预测性能由受试者工作特征曲线下面积(AUC)来评估。结果从脑血容量、脑血流量、平均传输时间和到达峰值时间灌注参数图中共提取影像组学特征4787个,去除冗余特征、进行特征筛选后共选取了 23个重要性高的影像组学特征进行最优模型拟合,其中以脑血容量参数特征或纹理特征对预测模型的贡献最大。经统计分析,分入训练集与测试集的患者之间的临床信息无明显差异[性别(P=0.2964)、年龄(P=0.3127)、WHO分级(P=0.6557)、分子亚型(P=0.9812)]。在训练集中,模型预测IDH野生型,IDH突变和1p/19q非共缺失型,IDH突变和lp/19q共缺失型的AUC值分别为0.831、0.743和0.764。在验证集中,IDH野生型的AUC值为0.783,IDH突变和1p/19q非共缺失型的AUC值为0.705,IDH突变和1p/19q共缺失型的AUC值为0.712。结论基于动态磁敏感灌注加权成像构建的影像组学模型可以提供一种可靠的分析方法来预测成人型弥漫性胶质瘤的分子分型,进而提前对胶质瘤患者进行早期干预,改善预后。更多还原

【Abstract】 BackgroundGlioma is one of the most common intracranial space-occupying lesions.Compared with other intracranial space-occupying lesions,such as meningioma,acoustic neuromas and ependymoma,primary glioma is highly invasive and destructive,with a high incidence,the most complicated treatment and extremely difficult to cure.Glioblastoma(WHO grade IV)was the most malignant,with a median survival time of about 15 months.In addition,genetic molecular markers in primary glioma are closely related to survival time,treatment feedback and prognostic status.Patients with IDH(isocitrate dehydrogenase)mutant glioma have relatively slow disease progression and long survival time.IDH wild type patients have rapid disease progression,poor prognosis and easy recurrence after surgical resection.The hypermethylation of MGMT(O6-methylguanine-DNA methyltransferase)promoter has been proved to be related to the chemical sensitivity of alkylating agent(temozolomide).TERT(telomerase reverse transcriptase)promoter mutations also indicates a shorter overall survival in patients with glioblastoma.In 2016,the World Health Organization(WHO)Central Nervous System(CNS)tumor classification included the IDH and chromosome 1p/19q into the classification features of gliomas,and classified adult diffuse gliomas into different statuses according to different status of IDH and 1p/19q.Three molecular subtypes:IDH wildtype(IDHwt),IDH mutant and 1p/19q non-codeletion(IDHmut-noncodel),IDH mutant and 1p/19q co-deletion(IDHmut-codel).Meanwhile,IDH and 1p/19q are considered as one of the important molecular markers for tumor classification,risk stratification and determination of postoperative treatment options in adult-type diffuse glioma.However,the information of various types of genetic molecular markers in tumor tissues needs to be confirmed by biopsy or histopathological diagnosis after surgical resection.Therefore,it is more important to find a non-invasive,preoperative and highly accurate method to predict tumor genetic molecular markers.In recent years,the classification,prognosis and survival of brain tumors have been deeply explored based on radiomics of convenient MR sequences(T1,T2,T1C,FLAIR,ADC,etc.).And the remarkable results have been gained.Dynamic sensitivity contrast perfusion weighted imaging(DSC-PWI)has obvious advantages in tumor vascular network and blood supply,which accords with the characteristics of dense blood vessels and strong neovascularization ability.Hence,we aim to establish a radiomics model based on dynamic sensitivity contrast perfusion weighted imaging extractions to predict three molecular subtypes of adult-type diffuse glioma,and then to intervene the tumor in advance to improve the survival and prognosis of patients.MethodsIn this retrospective study,from January 2013 to January 2020,1728 patients with glioma confirmed by histopathology after surgical resection were collected.278 cases of adult-type diffuse glioma were screened out.The cerebral blood volume,cerebral blood flow,mean transit time and time to peak maps were gotten from DSC-PWI.The radiomics features were extracted from the calibration image.After radiomic feature removal and feature selection,a radiomics model was constructed with the most relevant feature set,and the included patients were randomly divided into a training set(n=175)and a validation set(n=103)according to the ratio of 13:7.Finally,the prediction performance of the radiomics model was evaluated by determining the area under the receiver operating characteristic curve(AUC).ResultsA total of 4787 radiomic features were extracted from the maps of CBV,CBF,MTT,TTP.After removing redundant features and feature selecting,a total of 23 radiomic features were enrolled in optimal model simulation,in which the parameter features of CBV or texture features have the greatest contribution to the prediction model.No significant difference in clinical information was found between the glioma patients in the training group and the testing group(sex(P=0.2964),age(P=0.3127),WHO classification(P=0.6557),molecular subtype(P=0.9812)).In the training set,the prediction model achieved an AUC of 0.831,0.743 and 0.764 for IDHwt,IDHmut-noncode and IDHmut-codel,respectively.In the validation set,the model achieved an AUC of 0.783,0.705 and 0.712 for IDHwt,IDHmut-noncode and IDHmut-codel,respectively.ConclusionsRadiomics models based on dynamic sensitivity contrast perfusion weighted imaging can provide a reliable method to predict molecular subtypes of adult-type diffuse gliomas.And then early intervention for cancer patients can be carried out to improve the prognosis.更多还原