Lima Marcos Antonio Dias, Vasconcelos Carlos Frederico Motta, Ichinose Roberto Macoto, de Sá Antonio Mauricio Ferreira Leite Miranda
Instituto Alberto Luiz Coimbra de Pós-Graduação e Pesquisa de Engenharia - Universidade Federal do Rio de Janeiro (COPPE-UFRJ), Rio de Janeiro, RJ, Brazil.
Instituto Nacional de Câncer (INCA), Rio de Janeiro, RJ, Brazil.
Radiol Bras. 2025 May 20;58:e20240093. doi: 10.1590/0100-3984.2024.0093. eCollection 2025 Jan-Dec.
To develop a deep learning system to classify non-small cell lung cancer (NSCLC) by histologic subtype-adenocarcinoma or squamous cell carcinoma (SCC)-from computed tomography (CT) images in which the tumor regions were segmented, comparing our results with those of similar studies conducted in other countries and evaluating the accuracy of automated classification by using data from the Instituto Nacional de Câncer, Brazil.
To develop the classification system, we employed a 2D U-Net neural network for semantic segmentation, with data augmentation and preprocessing steps. It was pretrained on 28,506 CT images from The Cancer Image Archive, a private database, and validated on 2,015 of those images. To develop the classification algorithm, we used a VGG16-based network, modified for better performance, with 3,080 images of adenocarcinoma and SCC from the Instituto Nacional de Câncer database.
The algorithm achieved an accuracy of 84.5% for detecting adenocarcinoma and 89.6% for detecting SCC, with sensitivities of 91.7% and 90.4%, respectively, which are considered satisfactory when compared with the values obtained in similar studies.
The system developed appears to provide accurate automated detection, as well as tumor segmentation and classification of NSCLC subtypes of a local population using deep learning networks trained using public image data sets. This method could assist oncological radiologists by improving the efficiency of preliminary diagnoses.
开发一种深度学习系统，根据计算机断层扫描（CT）图像中分割出的肿瘤区域，对非小细胞肺癌（NSCLC）的组织学亚型——腺癌或鳞状细胞癌（SCC）进行分类，将我们的结果与其他国家开展的类似研究结果进行比较，并利用巴西国家癌症研究所的数据评估自动分类的准确性。
为开发分类系统，我们采用二维U-Net神经网络进行语义分割，并进行数据增强和预处理步骤。该网络在一个私人数据库——癌症图像存档库的28506张CT图像上进行预训练，并在其中2015张图像上进行验证。为开发分类算法，我们使用了一个基于VGG16的网络，并对其进行了改进以提高性能，使用了来自巴西国家癌症研究所数据库的3080张腺癌和SCC图像。
该算法检测腺癌的准确率为84.5%，检测SCC的准确率为89.6%，敏感性分别为91.7%和90.4%，与类似研究中获得的值相比，这些结果被认为是令人满意的。
所开发的系统似乎能够提供准确的自动检测，以及使用公共图像数据集训练的深度学习网络对当地人群的NSCLC亚型进行肿瘤分割和分类。这种方法可以通过提高初步诊断的效率来协助肿瘤放射科医生。
