深度学习方法在肺结节诊断中的应用

Application of Deep Learning Methods in Diagnosis of Lung Nodules

肺癌是世界上死亡率最高的恶性肿瘤疾病,对其进行早期诊断可以显著提高肺癌患者的生存率。深度学习能够提取医学图像的隐含层特征,并完成医学图像的分类及分割任务,因此应用深度学习方法实现肺结节的早期诊断任务成为研究热点。首先对肺结节诊断领域广泛使用的几个数据库进行介绍,然后结合近几年国内外发表的相关文献整理了深度学习框架应用于肺结节分割和分类的最新研究进展,总结并分析了各类算法的基本思想、网络架构形式、代表性改进方案以及优缺点等。最后讨论了深度学习在肺结节诊断过程中面临的一些问题并给出结论,并对发展前景进行了展望,以期为今后该领域的应用研究提供参考,从而加快该领域研究的成熟和临床落地应用。

Lung cancer is the malignant tumor with the highest mortality rate in the world.Its early diagnosis can remarkably improve the survival rate of lung cancer patients.Deep learning can extract the hidden layer features of medical images and can complete the classification and segmentation of medical images.The application of deep learning methods for the early diagnosis of lung nodules has become a key point of research.This article introduces several databases commonly used in the field of lung nodule diagnosis and combines the relevant literature recently published at home and abroad to classify the latest research progress and summarize and analyze the application of deep learning frameworks for lung nodule image segmentation and classification.The basic ideas of various algorithms,network architecture forms,representative improvement schemes,and a summary of advantages and disadvantages are presented.Finally,some problems encountered while using deep learning for the diagnosis of pulmonary nodules,conclusions,and the development prospects are discussed.This study is expected to provide a reference for future research applications and accelerate the maturity of research and clinical applications in the concerned field.