人工智能辅助CT肺结节良恶性鉴别的研究进展

陈师 胡剑鹏 徐伟 张骥 吴吉明

[摘要] 肺癌属于临床常见的恶性肿瘤之一，当前胸部CT是进行早期肺癌鉴别的重要方式，但是因其存在“异病同影”等情况，加之受到临床经验等因素影响，在病灶良恶性鉴别方面有较大差异，极易出现误诊或漏诊情况。近年来人工智能被逐渐应用于临床，其在肺结节良恶性鉴别方面也发挥着一定作用。本文从人工智能评估肺结节良恶性的基本过程、人工智能模型在鉴别肺结节良恶性方面效能、人工智能诊断肺结节效能的影响因素、问题及展望方面进行分析，以期提升人工智能辅助CT鉴别肺结节良恶性效果。

[关键词] 良恶性;肺结节;计算机断层扫描成像;人工智能;手动分割

[中图分类号] R445          [文献标识码] A          [文章编号] 1673-9701（2021）30-0184-04

[Abstract] Lung cancer is one of the common clinical malignant tumors. At present， chest CT is an important way to differentiate early lung cancer. However， because of its "different diseases with the same shadow" and the influence of clinical experience， there are significant differences in the differentiation of benign and malignant lesions， and it is easy to have misdiagnosis or missed diagnosis. In recent years， artificial intelligence has been gradually applied in clinical practice， which also plays a role in differentiating benign from malignant pulmonary tuberculosis. This article analyzes the basic process of artificial intelligence in evaluating benign and malignant pulmonary nodules， the efficacy of the artificial intelligence model in differentiating benign and malignant pulmonary nodules， and the influencing factors， problems and prospects of artificial intelligence in diagnosing pulmonary nodules， in order to improve the effect of artificial intelligence assisted CT in differentiating benign and malignant pulmonary nodules.

[Key words] Benign and malignant; Pulmonary nodules; Computed tomography imaging; Artificial intelligence; Manual segmentation

據一项临床调查显示[1]，男性肺癌患病率和死亡率约占全部恶性肿瘤首位，而女性肺癌患病率居于恶性肿瘤第三位，死亡率则仅次于乳腺癌。伴随科学技术水平的进步，人工智能技术被逐渐应用于临床中，其在鉴别疾病良恶性时能够发挥良好作用[2]。机器学习作为人工智能的关键技术，而深度学习计算方法作为新研发的一种方式，其展示形式为嵌套层次概念，能够发挥较强的功能性和灵活性[3]。对于传统机器学习而言，该种方式无需手动提取特征，仅需将病理检查情况和肺结节影像数据录入，则自动生成对应关系。所以，将其应用于高通量图像分析中，可发挥诊断价值高及效率快等特点。……