基于病理组织切片的肺腺癌肿瘤突变负荷预测模型

肺癌是目前死亡率最高的恶性癌症之一,其中非小细胞肺癌(NSCLC)致死率极高。最近医学研究发现,肿瘤突变负荷(TMB)对于癌症的免疫治疗和化疗的疗效具有较好的预测作用,但传统使用基因测序计算TMB的方法存在检测成本高、周期长、样本依赖度高等缺点。针对上述问题,本研究提出一种混合卷积神经网络和自注意力机制的深度学习模型(FCA-Former)用于预测TMB。该模型以CoAtNet为骨干网络,通过在网络中结合坐标注意力以及融合深度可分离卷积的方式,提高模型的运算速度以及对病理组织切片图像的全局特征提取能力。实验数据采用TCGA数据库中肺腺癌数字病理切片图像数据集,其中高TMB水平的样本271张,低TMB水平的样本66张。实验结果表明,所提方法达到的最高平均曲线下面积(AUC)为98.1%,比现有最好方法RcaNetr提高9.8%。此项研究结果对于NSCLC的预后治疗效果具有较强的指导意义。     

基于残差网络的肺癌肿瘤突变负荷多分类预测模型

经医学研究发现,肿瘤突变负荷（TMB）与非小细胞肺癌（NSCLC）免疫治疗的疗效呈正相关,并且TMB值对靶向治疗和化疗的疗效也有一定的预测作用。然而,计算TMB值需要借助全外显子组测序（WES）技术,成本较高。对此,本文利用临床常用的数字病理组织切片图像,研究TMB与切片图像之间的关联关系,并据此预测患者的TMB水平。本文提出了一种基于残差坐标注意力（RCA）结构并融合多尺度注意力引导（MSAG）模块的深度学习模型（RCA-MSAG）。该模型以50层残差网络（ResNet-50）为基准模型,并将坐标注意力(CA)融入到瓶颈（bottleneck）模块,用来捕获方向感知和位置敏感信息,从而使模型能够更准确定位和识别感兴趣的位置。然后,通过在网络内添加MSAG模块,使模型可以提取肺癌病理组织切片的深层特征以及通道之间的交互信息。本文实验数据集采用癌症基因组图谱（TCGA）公开数据集,数据集由200张肺腺癌病理组织切片组成,其中高TMB值的数据80张,中TMB值的数据77张,低TMB值的数据43张。实验结果表明,所提模型的准确率、精确率、召回率和F1分数分别为96.2%、96.4%、96.2...     

基于深度学习的肺腺癌肿瘤突变负荷的预测

肿瘤突变负荷(TMB)与非小细胞肺癌(NSCLC)的免疫治疗疗效呈正相关,并且在近期的相关研究中,肿瘤突变负荷对靶向治疗及化疗的疗效也有一定的预测作用.因此,提出一种融合注意力机制的Inception深度学习模型(CAIM),用于对TCGA数据库中的非小型细胞肺癌中的肺腺癌的病理切片进行识别.首先,对数据样本进行切分,...     

表皮生长因子受体在靶向治疗非小细胞肺癌中的研究进展

肺癌是最常见的恶性肿瘤之一,其发病率和死亡率呈逐年上升趋势,在全世界癌症死亡中占第一位[1]。按照临床和组织病理学特征,肺癌分为非小细胞肺癌(non-small cell lung cancers,NSCLC)和小细胞肺癌(small cell lung cancers,SCLC);其中,NSCLC约占80%,SCLC约占20%[2]。目前,NSCLC的治疗仍以化疗为主,常用的化疗药物为含     

非小细胞肺癌预后蛋白的研究进展

肺癌是一种常见的恶性肿瘤,其发病率、病死率呈逐年上升趋势.肺癌分为非小细胞肺癌(non-small cell lung cancer,NSCLC)和小细胞肺癌(small cell lung cancer,SCLC)2型,其中NSCLC约占所有肺癌的85%,其5年生存率约为15%.     

与肺癌相关的靶基因研究简介

肺癌是目前全世界发病率最高的恶性肿瘤，其病死率也居于前列。因此，肺癌的早期诊断及早期治疗对于降低肺癌患者的病死率以及改善患者生存质量至关重要。近年来有关肺癌早期诊断、治疗及预后的分子标志物的研究愈来愈受到关注。在肺癌中，非小细胞肺癌(non-small cell lung cancer，NSCLC)占绝大多数。针对晚期NSCLC，肿瘤分子靶向治疗日趋重要。因此，从分子水平研究肺癌的发病机制和诊断，对肺癌的防治有重大意义。近些年，对基因水平方面的研究也取得了很大进展。现对与肺癌相关的靶基因最新研究进行展综述如下。     

联合检测血清中CYFRA21-1、NSE对肺癌诊断的价值

目的探讨联合检测血清中肿瘤标记物CYFRA21-1、NSE对肺癌早期诊断、病情监测及疗效判定的价值。方法将肺癌患者分为非小细胞肺癌组(NSCLC)和小细胞肺癌组(SCLC),以健康人及肺部良性病变患者(BLD)为对照组。静脉取血,采用免疫放射法(IRMA)、放射免疫法(RIA)分别检测血清CYFRA21-1和NSE。结果血清CYFRA21-1水平在NSCLC组明显高于SCLC和BLD组(P<0.01),其对NSCLC诊断的阳性率显著高于SCLC组(58.1%vs16.7%,P<0.01)。NSE则在SCLC组明显高于NSCLC和BLD组(P<0.01),其对SCLC诊断的阳性率为66.7%,显著高于NSCLC组24.3%(P<0.01)。血清CYFRA21-1、NSE联合检测与各单项指标相比,可将肺癌诊断的敏感度、特异度和阳性率分别提高到79.1%、91.7%和85.0%(P<0.01)。此外,在NSCLC组,血清CYFRA21-1含量变化与肿瘤分期呈正相关,且与患者病情的变化有相关性。结论联合检测血清中CYFRA21-1、NSE水平可明显提高肺癌早期诊断的阳性率,从而提高肺癌的早期诊断。临床上动态监测血清CYFRA21-1水平对NSCLC的进程及疗效判断有一定的价值。     

联合检测血清中CYFRA21-1、NSE对肺癌诊断的价值

目的 探讨联合检测血清中肿瘤标记物CYFRA21-1、NSE对肺癌早期诊断、病情监测及疗效判定的价值.方法 将肺癌患者分为非小细胞肺癌组(NSCLC)和小细胞肺癌组(SCLC),以健康人及肺部良性病变患者(BLD)为对照组.静脉取血,采用免疫放射法(IRMA)、放射免疫法(RIA)分别检测血清CYFRA21-1和NSE.结果 血清CYFRA21-1水平在NSCLC组明显高于SCLC和BLD组(P＜0.01),其对NSCLC诊断的阳性率显著高于SCLC组(58.1%vs 16.7%,P＜0.01).NSE则在SCLC组明显高于NSCLC和BLD组(P＜0.01),其对SCLC诊断的阳性率为66.7%,显著高于NSCLC组24.3%(P＜0.01).血清CYFRA21-1、NSE联合检测与各单项指标相比,可将肺癌诊断的敏感度、特异度和阳性率分别提高到79.1%、91.7%和85.0%(P＜0.01).此外,在NSCLC组,血清CYFRA21-1含量变化与肿瘤分期呈正相关,且与患者病情的变化有相关性.结论 联合检测血清中CYFRA21-1、NSE水平可明显提高肺癌早期诊断的阳性率,从而提高肺癌的早期诊断.临床上动态监测血清CYFRA21-1水平对NSCLC的进程及疗效判断有一定的价值.     

肺癌与抑癌基因甲基化关系的研究进展

肺癌是当今世界发病率与死亡率位于前列的恶性肿瘤之一。肺癌一般分为两种：小细胞肺癌(SCLC)和非小细胞肺癌(NSCLC)。在临床中80％～85％为NSCLC，SCLC属于内分泌细胞来源的肿瘤，仅占20％～25％。过去人们对肺癌发生的研究主要局限于分析基因以点突变、小片段缺失、插入等方式造成的原癌基因的激活和抑癌基因的功能丧失。     

CA125、CEA、TSGF和VEGF联检在非小细胞肺癌诊治中的应用

目前,肺癌的发病率已居恶性肿瘤的首位,非小细胞肺癌(NSCLC)占肺癌的70～80%,且首诊时多已中、晚期,疗效不尽如人意,因此人们一直在积极寻找理想的肿瘤标志物,以期提高非小细胞肺癌的早期诊断率,并监测患者的疗效、复发及转移情况。为观察CA125、CEA、TSGF、血管内皮生长因子(V     

Tumor mutational burden assessed by targeted NGS predicts clinical benefit from immune checkpoint inhibitors in non-small cell lung cancer

In non-small cell lung cancer (NSCLC), immune checkpoint inhibitors (ICIs) significantly improve overall survival (OS). Tumor mutational burden (TMB) has emerged as a predictive biomarker for patients treated with ICIs. Here, we evaluated the predictive power of TMB measured by the Oncomine™ Tumor Mutational Load targeted sequencing assay in 76 NSCLC patients treated with ICIs. TMB was assessed retrospectively in 76 NSCLC patients receiving ICI therapy. Clinical data (RECIST 1.1) were collected and patients were classified as having either durable clinical benefit (DCB) or no durable benefit (NDB). Additionally, genetic alterations and PD-L1 expression were assessed and compared with TMB and response rate. TMB was significantly higher in patients with DCB than in patients with NDB (median TMB = 8.5 versus 6.0 mutations/Mb, Mann–Whitney p = 0.0244). 64% of patients with high TMB (cut-off = third tertile, TMB ≥ 9) were responders (DCB) compared to 33% and 29% of patients with intermediate and low TMB, respectively (cut-off = second and first tertile, TMB = 5–9 and TMB ≤ 4, respectively). TMB-high patients showed significantly longer progression-free survival (PFS) and OS (log-rank test p = 0.0014 for PFS and 0.0197 for OS). While identifying different subgroups of patients, combining PD-L1 expression and TMB increased the predictive power (from AUC 0.63 to AUC 0.65). Our results show that the TML panel is an effective tool to stratify patients for ICI treatment. A combination of biomarkers might maximize the predictive precision for patient stratification. Our study supports TMB evaluation through targeted NGS in NSCLC patient samples as a tool to predict response to ICI therapy. We offer recommendations for a reliable and cost-effective assessment of TMB in a routine diagnostic setting. © 2019 The Authors. The Journal of Pathology published by John Wiley & Sons Ltd on behalf of Pathological Society of Great Britain and Ireland.     

基于深度级联网络的乳腺淋巴结全景图像的癌转移区域自动识别

淋巴结癌转移区域的自动识别是乳腺癌病理分期的重要前提。但由于全景图像尺寸巨大, 组织形态复杂多样, 在乳腺淋巴结全景图像中自动检测和定位癌转移区域具有很大的难度。设计一种基于深度级联网络的方法, 实现对乳腺淋巴结全景图像癌转移区域的自动定位与识别。采用由粗定位到精定位的两个深度网络模型级联的方式, 首先基于医生标记的癌转移区域, 提取阳性与阴性图像块训练粗定位网络VGG16得到粗定位结果, 然后对比粗定位结果与医生标记提取阳性和假阳性区域的图像块, 再训练精定位的ResNet50网络用于识别阳性和假阳性区域。为了验证所提出深度级联网络的有效性, 选用Camelyon16公开的共400张乳腺淋巴结全景图像数据集用作训练和测试。结果表明, 所提出的VGG16+ResNet50级联网络模型的定位指标FROC得分达到0.891 2, 分别比单个深度网络模型VGG16和ResNet50的FROC得分高0.153 1和0.147 0, 比AlexNet+VGG16级联的网络模型FROC得分高0.028 8, 显示深度级联网络模型对淋巴结癌转移区域可以实现更加精准的识别。     

The role of molecular analyses in the diagnosis and treatment of non-small-cell lung carcinomas

Non-small-cell lung cancer (NSCLC) subtyping has recently been a key factor in determining patient management with novel drugs. In addition, the identification of distinct oncogenic driver mutations frequently associated with NSCLC histotype and coupled to the clinical responses to targeted therapies have revolutionized the impact of histologic type and molecular biomarkers in lung cancer. Several molecular alterations involving different genes (EGFR, KRAS, ALK, BRAF, and HER2) seem to have a remarkable predilection for adenocarcinoma and specific inhibitors of EGFR and ALK are now available for patients with adenocarcinoma harboring the relevant gene alterations. The efficacy of histology-based and molecular-targeted therapies had a deep impact in (1) re-defining classification of lung cancer (particularly adenocarcinomas) and (2) routine clinical practice of pathologists involved in optimization of handling of tissue samples in order to guarantee NSCLC subtyping with the help of immunohistochemistry and adequately preserve tumor cells for molecular analysis. In agreement with the modern multidisciplinary approach to lung cancer, we reviewed here the diagnostic and predictive value of molecular biomarkers according to the clinical, pathologic, and molecular biologist viewpoints.     

基于深度学习结合解剖学注意力机制的肺结节良恶性分类

肺结节作为肺癌的初期表现,及时的发现和准确的良恶性诊断对于疾病的治疗具有重要的意义。为了提高肺部CT图像中肺结节良恶性的诊断率,提出一种基于3D ResNet的卷积神经网络,并通过加入解剖学注意力模块有效地提高了肺结节良恶性的分类精度。此外,该方法通过自动分割以获取注意力机制所需的感兴趣区域,实现整个流程的全自动化。解剖学注意力的添加能更好地捕捉图像中的局部纹理信息,进一步提取对于肺结节良恶性诊断有用的特征。本文方法在LIDC-IDRI数据集上进行验证。实验结果表明与传统的3D ResNet及其他现有的方法相比,本文方法在分类精度上有显著的提高,在独立测试集上的最终分类的AUC达到0.973,准确率为0.940。由此可见,本文方法能在辅助医生对肺结节的诊断中起到重要作用。     

Screening and identification of potential novel lipid biomarkers for non-small cell lung cancer using ultra-high performance liquid chromatography tandem mass spectrometry

Lung cancer is characterized by a high incidence rate and low survival rate. It is important to achieve early diagnosis of the disease. We applied ultra-high performance liquid chromatography tandem mass spectrometry to screen plasma lipid spectrum in non-small cell lung cancer (NSCLC) patients, healthy controls (HC), and community-acquired pneumonia (CAP) patients. Modeling employing orthogonal partial least squares-discriminant analysis combined with t-test was used to screen the differential lipids. Logistic regression analysis was used to establish the diagnostic model, while the accuracy was verified by 10-fold cross-validation. The results showed that the abnormal metabolism of lipid in NSCLC mainly comprised fatty acid metabolism, phospholipid metabolism, and glyceride metabolism. Four potential biomarkers, including LPC (14:0/0:0), LPI (14:1/0:0), DG (14:0/18:2/0:0), and LPC (16:1/0:0), were fitted by the receiver operating characteristic curve model with the area under curve (AUC) value of 0.856, and the specificity and sensitivity were 87.0 and 78.0%, respectively. The results of cross validation showed that the AUC value of the model was 0.812, the sensitivity was 72.9%, and the specificity was 82.6%. The positive rate of four potential lipid biomarkers in this study (>60.0%) was higher than that of existing tumor biomarkers in the clinical application. We investigated the plasma lipid profile of NSCLC patients and identified lipid biomarkers with potential diagnostic values. From the lipidomics perspective, our study may lay a foundation for the biomarker-based early diagnosis of lung cancer.     

细胞角蛋白片段19等指标在非小细胞肺癌诊断中的应用

目的　研究胸水和血清细胞角蛋白片段 19(CK - 19)等指标在非小细胞性肺癌 (NSCLC)实验室诊断中的应用价值 .方法　采用ELISA法检测 4 5例NSCLC、5 0例良性肺疾病患者血清和胸水中CK - 19水平 ,30例健康体检者血清CK - 19水平 ,同步测定了乳酸脱氢酶 (LDH)、腺苷脱氨酶 (ADA)、C反应蛋白 (CRP)、癌胚抗原 (CEA)、免疫球蛋白E(IgE)水平 ,进行显著性检验及相关性分析 .结果　NSCLC组血清和胸水中CK - 19、CEA和LDH水平均明显高于良性肺疾病组 (p <0 .0 5 ) ,而血清ADA水平均低于良性肺疾病组和健康对照组 (p <0 .0 5 ) .良性肺疾病组血清和胸水CRP水平明显高于NSCLC组 (p <0 .0 5 ) .胸水中CK - 19水平与血清CK - 19水平高度相关 (NSCLC组r=0 .86 8;良性肺疾病组r=0 .5 4 6 2 ) .胸水和血清CK - 19对肺癌的诊断敏感性、特异性和准确性分别为 93.3%和 71.1%、91.7%和 87.5 %、92 %和 81.6 % .结论　胸水和血清CK - 19检测对NSCLC的诊断是一个较好的指标 ,联合检测相关指标有助于临床诊断     

联合检测7种自身抗体在非小细胞肺癌中的诊断价值

目的 探讨7种肿瘤相关抗原自身抗体在非小细胞肺癌诊断中的应用价值.方法 回顾性分析非小细胞肺癌患者443例、体检健康人群405例,采用ELISA法检测各组血清中7种自身抗体的水平,分析抗体水平组间及阳性率差异,绘制ROC曲线分析单个抗体和联合检测的诊断效能,结合病理诊断结果 分析联合检测的阳性率与肺癌临床病理特征的关系...     

小细胞肺癌肿瘤标志物：血清NSE与CYFRA21—1和CEA

目的　评价小细胞肺癌 (SCLC)肿瘤标志物血清神经特异性烯醇化酶 (NSE)的诊断价值 ,并与细胞角质蛋白 (CYFRA2 1- 1) ,癌胚抗原 (CEA)进行比较 .方法　采用放射性免疫学方法和化学发光法测定 2 4例SCLC ,92例非小细胞肺癌 (NSCLC)和 118例肺良性疾病患者血清NSE ,CYFRA2 1- 1和CEA .结果　血清NSE对SCLC患者敏感性、特异性和准确率分别为 87.5 %、85 .7%和 85 .0‰ ,SCLC患者血清NSE水平高于NSCLC(p<0 .0 5 ) ,SCLC患者血清NSE灵敏度明显高于CYFRA2 1- 1和CEA(p均 <0 .0 1) .NSE与CYFRA2 1- 1或CEA联合可提高敏感性 .血清NSE水平在SCLC患者化疗前高于正常 ,化疗后 3周升高明显 ,缓解后降至正常 .结论　NSE是一种有效SCLC肿瘤标志物 ,与CYFRA2 1- 1或CEA联合可提高敏感性 ,有效性 ,血清NSE水平还可精确地反映SCLC患者病情变化     

Residual Deep Convolutional Neural Network Predicts MGMT Methylation Status

Predicting methylation of the O6-methylguanine methyltransferase (MGMT) gene status utilizing MRI imaging is of high importance since it is a predictor of response and prognosis in brain tumors. In this study, we compare three different residual deep neural network (ResNet) architectures to evaluate their ability in predicting MGMT methylation status without the need for a distinct tumor segmentation step. We found that the ResNet50 (50 layers) architecture was the best performing model, achieving an accuracy of 94.90% (+/? 3.92%) for the test set (classification of a slice as no tumor, methylated MGMT, or non-methylated). ResNet34 (34 layers) achieved 80.72% (+/? 13.61%) while ResNet18 (18 layers) accuracy was 76.75% (+/? 20.67%). ResNet50 performance was statistically significantly better than both ResNet18 and ResNet34 architectures (p < 0.001). We report a method that alleviates the need of extensive preprocessing and acts as a proof of concept that deep neural architectures can be used to predict molecular biomarkers from routine medical images.