arXivDaily arXiv每日学术速递 周一至周五更新
arXiv周末暂无论文更新,休息一下吧,周末愉快~~
arXiv 2104.14420cs.CVcs.LG
收藏

基因型导向放射组学标记预测非小细胞肺癌复发

Genotype-Guided Radiomics Signatures for Recurrence Prediction of Non-Small-Cell Lung Cancer

Ritsumeikan University(立命馆大学);Zhejiang University(浙江大学);Zhejiang Lab(之江实验室)

Panyanat Aonpong, Yutaro Iwamoto, Xian-Hua Han, Lanfen Lin, Yen-Wei Chen

摘要:

非小细胞肺癌(Non-small cell lung cancer,NSCLC)是一种严重的疾病,手术后复发率高。最近,已经提出了许多机器学习方法用于递归预测。使用基因数据的方法具有较高的预测精度,但需要较高的成本。虽然仅使用CT图像的放射组学签名不昂贵,但其准确性相对较低。在本文中,我们提出了一种基因型引导放射组学方法(GGR),以获得高的预测精度,低成本。我们使用了NSCLC的公共放射基因组学数据集,其中包括CT图像和基因数据。所提出的方法是一个两步的方法,它包括两个模型。第一个模型是基因估计模型,用于从计算机断层扫描(CT)图像中提取的放射组学特征和深度特征估计基因表达。第二个模型用于使用估计的基因表达数据来预测复发。本文提出的GGR方法是基于混合特征设计的,它是基于手工操作和基于深度学习的结合。实验结果表明,GGR算法的预测准确率从78.61%(现有放射组学方法)和79.14%(深度学习方法)提高到83.28%,具有显著的统计学意义。

英文摘要:

Non-small cell lung cancer (NSCLC) is a serious disease and has a high recurrence rate after the surgery. Recently, many machine learning methods have been proposed for recurrence prediction. The methods using gene data have high prediction accuracy but require high cost. Although the radiomics signatures using only CT image are not expensive, its accuracy is relatively low. In this paper, we propose a genotype-guided radiomics method (GGR) for obtaining high prediction accuracy with low cost. We used a public radiogenomics dataset of NSCLC, which includes CT images and gene data. The proposed method is a two-step method, which consists of two models. The first model is a gene estimation model, which is used to estimate the gene expression from radiomics features and deep features extracted from computer tomography (CT) image. The second model is used to predict the recurrence using the estimated gene expression data. The proposed GGR method designed based on hybrid features which is combination of handcrafted-based and deep learning-based. The experiments demonstrated that the prediction accuracy can be improved significantly from 78.61% (existing radiomics method) and 79.14% (deep learning method) to 83.28% by the proposed GGR.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑