Fig. 1From: The diagnostic and prognostic value of radiomics and deep learning technologies for patients with solid pulmonary nodules in chest CT imagesThe framework of the deep learning model. A The pre-trained model. B The proposed model. “1 × 1 × 1 Conv” refers to the convolutional layer with 1 × 1 × 1 kernel. Dice loss refers to Dice similarity coefficient between the nodule mask and class activation mapping (attention map)Back to article page