基于影像组学特征的恶性孤立性肺结节诊断模型的开发
Development of a diagnostic model for malignant solitary pulmonary nodules based on radiomics features.
作者信息
Zhao Wei, Zou Chenxi, Li Chunsun, Li Jie, Wang Zirui, Chen Liang'an
机构信息
Department of Respiratory and Critical Care Medicine, General Hospital of the People's Liberation Army, Beijing, China.
Department of Pathology, General Hospital of the People's Liberation Army, Beijing, China.
出版信息
Ann Transl Med. 2022 Feb;10(4):201. doi: 10.21037/atm-22-462.
BACKGROUND
This study proposed a precise diagnostic model for malignant solitary pulmonary nodules (SPNs). This model can be used to identify objective and quantifiable image features and guide the clinical treatment strategy adopted for SPNs. This model will help clinicians optimize management strategies for SPN.
METHODS
In this retrospective study, the clinical data of 455 patients of SPN with defined pathological diagnosis between September 2016 and August 2019 were collected and analyzed. The data included pathological diagnosis, preoperative computed tomography (CT) diagnosis, gender, age, smoking history, family history of tumor, previous history, and contact history data. The quantitative image features and radiomic information of the SPNs were provided using computer-aided detection (CAD) "digital lung" software. The Chi-squared test was used to assess the accuracy between CAD and conventional CT in the diagnosis of SPNs. The diagnostic model for benign or malignant SPNs was developed using a multivariate logistic regression analysis that comprises 6 radiomic factors (irregularity, average diameter, COPD910, proportion of emphysema, proportion of fat, and average density of related blood vessels). The area under the receiver operating characteristic curve was used to evaluate the performance of the model in determining SPN risk of malignancy.
RESULTS
There was a statistical difference in the accuracy of CAD and conventional CT in diagnosing SPNs. According to the golden standard pathological diagnosis, the diagnostic accuracy of CAD (81%) was higher than that of conventional CT (63.7%) (P<0.05). Six variables (i.e., irregularity, the mean diameter, COPD910, the proportion of emphysema, the proportion of fat, and the vascular density) were identified using multivariable logistic regression to establish the diagnostic model for distinguish benign or malignant SPNs. The area under the receiver operating characteristic (ROC) curve (AUC) of the diagnostic model was 0.876 (95% CI: 0.8445-0.9076), and its sensitivity and specificity were 81.25% and 82.56% respectively.
CONCLUSIONS
The proposed diagnostic model, which comprises 6 radiomic factors, is accurate and effective at diagnosing benign or malignant SPNs.
背景
本研究提出了一种用于恶性孤立性肺结节（SPN）的精确诊断模型。该模型可用于识别客观且可量化的图像特征，并指导针对SPN采取的临床治疗策略。该模型将有助于临床医生优化SPN的管理策略。
方法
在这项回顾性研究中，收集并分析了2016年9月至2019年8月期间455例具有明确病理诊断的SPN患者的临床数据。数据包括病理诊断、术前计算机断层扫描（CT）诊断、性别、年龄、吸烟史、肿瘤家族史、既往史和接触史数据。使用计算机辅助检测（CAD）“数字肺”软件提供SPN的定量图像特征和放射组学信息。采用卡方检验评估CAD与传统CT在SPN诊断中的准确性。使用包含6个放射组学因素（不规则性、平均直径、COPD910、肺气肿比例、脂肪比例和相关血管平均密度）的多变量逻辑回归分析建立良性或恶性SPN的诊断模型。采用受试者操作特征曲线下面积评估该模型在确定SPN恶性风险方面的性能。
结果
CAD与传统CT在SPN诊断准确性方面存在统计学差异。根据金标准病理诊断，CAD的诊断准确性（81%）高于传统CT（63.7%）（P<0.05）。使用多变量逻辑回归确定了6个变量（即不规则性、平均直径、COPD910、肺气肿比例、脂肪比例和血管密度），以建立区分良性或恶性SPN的诊断模型。诊断模型的受试者操作特征（ROC）曲线下面积（AUC）为0.876（95%CI：0.8445 - 0.9076），其敏感性和特异性分别为81.25%和82.56%。
结论
所提出的包含6个放射组学因素的诊断模型在诊断良性或恶性SPN方面准确有效。
Zotero 插件