Department of Nuclear Medicine, Soonchunhyang University Hospital, Seoul, Republic of Korea.
Department of Pathology, Ajou University School of Medicine, Suwon, Republic of Korea.
Anticancer Res. 2022 Dec;42(12):5875-5884. doi: 10.21873/anticanres.16096.
BACKGROUND/AIM: We explored the prediction of programmed cell death ligand 1 (PD-L1) expression level in non-small cell lung cancer using a machine learning approach with positron emission tomography/computed tomography (PET/CT)-based radiomics.
A total of 312 patients (189 adenocarcinomas, 123 squamous cell carcinomas) who underwent F-18 fluorodeoxyglucose PET/CT were retrospectively analysed. Imaging biomarkers with 46 CT and 48 PET radiomic features were extracted from segmented tumours on PET and CT images using the LIFEx package. Radiomic features were ranked, and the top five best feature subsets were selected using the Gini index based on associations with PD-L1 expression in at least 50% of tumour cells. The areas under the receiver operating characteristic curves (AUCs) of binary classifications afforded by several machine learning algorithms (random forest, neural network, Naïve Bayes, logistic regression, adaptive boosting, stochastic gradient descent, support vector machine) were compared. The model performances were tested by 10-fold cross validation.
We developed and validated a PET/CT-based radiomic model predicting PD-L1 expression levels in lung cancer. Long run high grey-level emphasis, homogeneity, mean Hounsfield unit, long run emphasis from CT, and maximum standardised uptake value from PET were the five best feature subsets for positive PD-L1 expression. The Naïve Bayes model (AUC=0.712), with a sensitivity of 75.3% and specificity of 58.2%, outperformed all other classifiers. It was followed by the neural network model (AUC=0.711), random forest (AUC=0.700), logistic regression (AUC=0.673) and adaptive boosting (AUC=0.604).
PET/CT-based radiomic features may help clinicians identify tumours with positive PD-L1 expression in a non-invasive manner using machine learning algorithms.
背景/目的：我们使用基于正电子发射断层扫描/计算机断层扫描（PET/CT）的放射组学的机器学习方法探索了非小细胞肺癌中程序性死亡配体 1（PD-L1）表达水平的预测。
回顾性分析了 312 名接受 F-18 氟脱氧葡萄糖 PET/CT 的患者（189 例腺癌，123 例鳞状细胞癌）。使用 LIFEx 包从 PET 和 CT 图像上的分割肿瘤中提取了 46 个 CT 和 48 个 PET 放射组学特征。使用基于基尼指数的方法，根据与至少 50%肿瘤细胞中 PD-L1 表达的相关性，对放射组学特征进行排序，并选择前五个最佳特征子集。比较了几种机器学习算法（随机森林、神经网络、朴素贝叶斯、逻辑回归、自适应增强、随机梯度下降、支持向量机）提供的二进制分类的受试者工作特征曲线（AUC）的面积。通过 10 倍交叉验证测试模型性能。
我们开发并验证了一种基于 PET/CT 的放射组学模型，用于预测肺癌中的 PD-L1 表达水平。长运行高灰度强调、同质性、平均亨斯菲尔德单位、从 CT 上的长运行强调和从 PET 上的最大标准化摄取值是用于阳性 PD-L1 表达的五个最佳特征子集。朴素贝叶斯模型（AUC=0.712），灵敏度为 75.3%，特异性为 58.2%，优于所有其他分类器。其次是神经网络模型（AUC=0.711）、随机森林（AUC=0.700）、逻辑回归（AUC=0.673）和自适应增强（AUC=0.604）。
基于 PET/CT 的放射组学特征可以帮助临床医生使用机器学习算法以非侵入性的方式识别具有阳性 PD-L1 表达的肿瘤。
