肺癌预测模型及其进展
Lung Cancer Prediction Model and Its Progression

肺癌是我国及世界范围内发病率和死亡率最高的恶性肿瘤之一。随着低剂量计算机断层扫描的普及，我国肺癌的检出率逐年升高，全球癌症生存趋势监测报告数据显示，2000~2014年间肺癌的年龄标化5年生存率波动在10%~20%之间，在中国，肺癌的年龄标化5年生存率在2003~2015年间略有上升的趋势，但仍低于20.0%；2012~2015年中国人群肺癌的5年生存率仅为19.7%。影响肺癌5年生存率的关键因素在于临床诊断时的肿瘤分期。临床研究数据显示，早期肺癌的5年生存率可达61.2%，其中小于1 cm的I期肺癌的5年生存率达92%；而晚期肺癌的5年生存率仅7.0%。因此，早期诊断对于改善肺癌患者的预后至关重要。然而，中国肺癌病例的诊断以晚期居多，早期诊断率较低。预计至2025年，我国每年新发的肺癌患者将达到100万，其中约有75%的肺癌患者就诊时已属晚期，形成了发病率高、病死率高、5年存活率低的“二高一低”特点。肺癌发病时间较短且致病原因复杂，可在短时间内病灶转移，早期阶段以肺结节的形式存在，常因为其隐匿性而错过了最佳的治疗时间，而肺癌相关的预测模型的研究也在近些年较为热门，因为其有助于帮助患者早期诊断肺部结节的良恶倾向，本文主要讨论肺癌相关预测模型的应用和其和人工智能相结合的发展。
Lung cancer is one of the most common malignant tumors with the highest morbidity and mortality in China and worldwide. With the popularity of low-dose computed tomography, the detection rate of lung cancer in China has increased year by year. According to the Global Cancer Survival Trend Monitoring Report, the age-standardized 5-year survival rate of lung cancer fluctuated between 10% and 20% from 2000 to 2014. In China, the age-standardized 5-year survival rate of lung cancer showed a slight upward trend from 2003 to 2015, but it was still lower than 20.0%; the 5-year sur-vival rate of lung cancer in the Chinese population was only 19.7% from 2012 to 2015. The key fac-tor affecting the 5-year survival rate of lung cancer is the tumor stage at the time of clinical diagno-sis. Clinical research data show that the 5-year survival rate of early lung cancer can reach 61.2%, among which the 5-year survival rate of stage I lung cancer with a size less than 1 cm is 92%, while the 5-year survival rate of stage III lung cancer is only 7.0%. Therefore, early diagnosis is crucial for improving the prognosis of lung cancer patients. However, the diagnosis of lung cancer cases in Chi-na is mainly in late stage, and the early diagnosis rate is low. It is estimated that by 2025, the num-ber of new lung cancer patients in China will reach 1 million annually, and about 75% of them are in late stage at the time of diagnosis, forming the characteristics of high morbidity, high mortality, and low 5-year survival rate. Lung cancer has a short onset time and complex pathogenic factors, and the lesions can metastasize in a short time. In the early stage, they exist in the form of pulmonary nodules, which often miss the best treatment time because of their insidiousness. In recent years, the research on lung cancer-related prediction models has become popular, because they help pa-tients diagnose the benign and malignant tendency of pulmonary nodules in an early stage. This paper mainly discusses the application and development of