Lung cancer screening efficacy enhanced through radiomic and epigenetic biomarkers

摘要 吸烟是肺癌和阻塞性肺病的最大危险因素。国家之肺 筛选试验（NLST）入组了报告吸烟史至少为30包年的受试者， 研究表明，每年一次的低剂量计算机断层扫描（LDCT）筛查可以降低肺癌的死亡率， 与传统的胸部X光相比，大约减少了16%。然而，在临床上， 有效地区分少量恶性结节和许多良性肺结节， 筛选此外，在筛选期间捕获的胸部LDCT数据在定量评估中也具有未开发的实用性。 评估阻塞性肺疾病。 LDCT捕获了大量的信息，可以自动客观地量化和提取 使用计算机算法从图像数据中。我们有自动分割结构的方法， 感兴趣的图像数据，并将提取数百个放射性生物标志物集中在肺结节， 结节周围肺实质、全肺和捕获肺叶异质性。这项研究还将纳入 通过测量cg 05575921处的DNA甲基化，获得吸烟史的客观表观遗传生物标志物。我们 几项研究表明，表观遗传生物标志物可有力地预测吸烟强度。我们将使用 客观的放射学和表观遗传学生物标志物和机器学习方法来预测（1） 肺癌和（2）NLST筛查人群中阻塞性肺病快速进展。我们 假设在cg 05575921处引入DNA甲基化将是对这两种预测有价值的补充 模型确定假设的结果将指导是否应纳入这种表观遗传生物标志物 在前瞻性肺癌筛查研究中。 该项目将产生影响，因为它将导致改进的自动风险预测算法，以指导 在通过LDCT筛查检测到肺结节的受试者中进行管理。这种方法可以帮助快速 为癌症患者提供治疗，并预防侵入性诊断测试的并发症， 良性病变患者的诊断成像不必要的辐射暴露。快速预测 阻塞性肺疾病进展可能有利于临床医生/受试者共同决策讨论 和有针对性的戒烟干预措施，除了改善肺癌的预测。

ABSTRACT Smoking is the largest risk factor for both lung cancer and obstructive lung disease. The National Lung Screening Trial (NLST) enrolled subjects who reported a cigarette smoking history of at least 30 pack years and showed that annual low-dose computed tomography (LDCT) screening could reduce mortality from lung cancer by approximately 16%, compared to conventional chest x-ray. However, it remains clinically challenging to efficiently distinguish the small number of malignant nodules from the many benign lung nodules detected with screening. In addition, the chest LDCT data captured during screening also has untapped utility in quantitatively evaluating obstructive lung disease. LDCT captures a wealth of information that can be automatically and objectively quantified and extracted from the image data using computer algorithms. We have methods for automated segmentation of structures of interest from the image data and will extract hundreds of radiological biomarkers focused on pulmonary nodules, peri-nodular lung parenchyma, the whole lung, and capture lobar heterogeneity. This study will also incorporate an objective epigenetic biomarker of smoking history via measurement of DNA methylation at cg05575921. Our epigenetic biomarker has been shown to strongly predict smoking intensity by several studies. We will use the objective radiological and epigenetic biomarkers and machine learning approaches to predict both (1) the risk of lung cancer and (2) rapid obstructive lung disease progression in the NLST screening population. We hypothesize that incorporating DNA methylation at cg05575921 will be a valuable addition to both prediction models. Determining the outcome of the hypothesis will guide if this epigenetic biomarker should be incorporated in prospective lung cancer screening studies. This project will have impact as it will result in an improved automatic risk prediction algorithm to guide management in subjects with a lung nodule detected by LDCT screening. This approach can facilitate rapid treatment for those with cancer and prevent complications from invasive diagnostic testing as well as unnecessary radiation exposure from diagnostic imaging in those with benign lesions. Predicting rapid obstructive lung disease progression may be beneficial for clinician/subject shared decision-making discussions and targeted smoking cessation interventions in addition to improving lung cancer prediction.