深度学习在肺癌^(18)F-FDG PET/CT诊疗中的应用研究进展

Advance of deep learning in diagnosis and treatment of^(18)F-FDG PET/CT for lung cancer

肺癌是中国常见的恶性肿瘤,其发病率及死亡率均居癌症首位。由于早期症状不典型且缺乏有效的筛查手段,大多数肺癌患者发现时已处于晚期,预后较差。改善预后的关键在于早发现、早诊断、早治疗。随着人工智能和医疗大数据分析等前沿技术的不断进步,深度学习作为目前人工智能发展迅猛的一大分支,被认为是医学图像分析领域的宝贵工具,现已在早期肺癌的筛查、诊疗和预后评估以及晚期肺癌的随访中广泛应用,并取得诸多成果。本文阐述了近年来深度学习在肺癌^(18)F-FDG正电子发射体层成像(positron emission tomography,PET)/计算机体层成像(computed tomography,CT)诊疗中的应用现状及研究进展,主要介绍图像采集与重建、病灶检测与分割、诊断与鉴别诊断、基因突变状态与免疫治疗靶点及治疗反应与结局预测等,并对其发展前景及面临的挑战予以梳理。

Lung cancer is the common malignant tumor in China,and its incidence and mortality rank first.Due to atypical early symptoms and lack of effective screening methods,most lung cancer patients are detected at an advanced stage and often lead to poor prognosis.The key factors for improving the prognosis of lung cancer patients are early detection,early diagnosis and early treatment.With the continuous progress of cutting-edge technologies such as artificial intelligence and medical big data analysis,Deep learning,as a major branch of artificial intelligence,is recognized as a valuable tool in the field of medical image analysis,and has been widely used in the screening and diagnosis of early-stage lung cancer,treatment decision-making,prediction of prognosis and follow-up of advanced lung cancer,and many results have been reported in previous literatures.This paper reviewed the development status of deep learning technology in the diagnosis and treatment of lung cancer^(18)F-FDG PET/CT in recent years,covered the aspects of image acquisition and reconstruction,lesion detection and segmentation,diagnosis and differential diagnosis,gene mutation status and molecular therapeutic target prediction,treatment response and outcome prediction,and analyzed its development prospects and challenges.