摘要
通过表皮生长因子受体(Epidermal Growth Factor Receptor,EGFR)突变情况可以对患者是否患有非小细胞肺癌(Non-small Cell Lung Cancer,NSCLC)进行检测。提出了一种基于对比学习的自监督EGFR基因突变预测方法,在不需要大量专家手工标注患者数据集的情况下,对输入网络的患者病灶区图像进行阴性、阳性预测。对自监督BYOL网络进行修改,增加了网络投影层非线性多层感知器(Multilayer Perceptron,MLP)的层数,并将患者CT和PET两个模态的图像数据融合作为网络的输入,在不需要大量标注患者数据集的情况下,对阴性、阳性病例进行预测。在非小细胞肺癌EGFR基因突变数据集上,与传统的影像组学、有监督VGG-16网络、有监督ResNet-50、有监督Inception v3和无监督迁移学习CAE进行对比。实验结果表明,使用对比学习从患者的CT和PET图像学习到的患者病灶区图像的实例特征可以对阴性、阳性病例进行区分,并取得了77%的曲线下面积(Area Under the Curve,AUC);相对于传统的影像组学方法分类结果AUC提高了7%,相对于有监督VGG-16网络的分类结果AUC提高了5%;在不需要大量专家手工标注数据集及大量患者临床数据的情况下仅比有监督ResNet-50 AUC低9%。改进BYOL网络仅需要少量标注的患者数据集便可得到比部分传统有监督方法更准确的检测结果,展示了其辅助临床决策的潜力。
The epidermal growth factor receptor(EGFR)mutation status can predict whether a patient has non-small cell lung cancer(NSCLC).A self-supervised EGFR gene mutation prediction method based on contrastive learning is proposed,which can distinguish between negative and positive images of the patient’s lesion area input to the network,without requiring a large number of expert hand-labeled patient datasets.The self-supervised BYOL network was modified to increase the number of layers of the non-linear multilayer perceptron(MLP)of the network projection layer,and image data of the patient's CT and PET modalities were merged as the input of the network.Negative and positive medical records can be predicted without the need to annotate a large number of patient datasets.Using the non-small cell lung cancer EGFR gene mutation datasets,it is compared with traditional radiomics,supervised VGG-16 network,supervised ResNet-50 network,supervised Inception v3 network,and unsupervised transfer learning CAE.The experimental results show that the instance features of patient lesion area images learned from CT and PET images of patients using contrastive learning can be used to distinguish negative and positive cases,with an area under the curve(AUC)of 77%.The classification results improved by AUC of 7%compared to the traditional radiomics method,and by AUC of 5%compared to the classification results of the supervised VGG-16 network.The AUC is only 9%lower than that of supervised ResNet-50,without requiring a large number of expert hand-annotated datasets and large patient clinical datasets.The improved BYOL network proposed in this paper requires only a small number of labeled patient datasets to obtain more accurate prediction results than some traditional supervised methods,demonstrating its potential to help clinical decision-making.
作者
杨嘉楠
王忠昊
王昊霖
耿国华
曹欣
YANG Jianan;WANG Zhonghao;WANG Haolin;GENG Guohua;CAO Xin(College of Information Science and Technology,Northwest University,Xi'an 710127,China;National and Local Joint Engineering Research Center for Cultural Heritage Digitization,Northwest University,Xi'an 710127,China)
出处
《光学精密工程》
EI
CAS
CSCD
北大核心
2022年第9期1080-1090,共11页
Optics and Precision Engineering
基金
国家自然科学基金资助项目(No.61701403)
中国博士后科学研究基金资助项目(No.2018M643719)
陕西省科协青年人才托举计划资助项目(No.20190107)。
关键词
医学图像处理
深度学习
对比学习
PET/CT
肺非小细胞肺癌基因突变预测
medical image processing
deep learning
contrastive learning
PET/CT
prediction of gene mutations in lung non-small cell lung cancer
