Wang Lu, Zhou He, Xu Nan, Liu Yuchan, Jiang Xiran, Li Shu, Feng Chaolu, Xu Hainan, Deng Kexue, Song Jiangdian
Department of Library, Shengjing Hospital of China Medical University, Shenyang, Liaoning 110004, China.
School of Health Management, China Medical University, Shenyang, Liaoning 110122, China.
iScience. 2023 May 30;26(7):107005. doi: 10.1016/j.isci.2023.107005. eCollection 2023 Jul 21.
Proposing a general segmentation approach for lung lesions, including pulmonary nodules, pneumonia, and tuberculosis, in CT images will improve efficiency in radiology. However, the performance of generative adversarial networks is hampered by the limited availability of annotated samples and the catastrophic forgetting of the discriminator, whereas the universality of traditional morphology-based methods is insufficient for segmenting diverse lung lesions. A cascaded dual-attention network with a context-aware pyramid feature extraction module was designed to address these challenges. A self-supervised rotation loss was designed to mitigate discriminator forgetting. The proposed model achieved Dice coefficients of 70.92, 73.55, and 68.52% on multi-center pneumonia, lung nodule, and tuberculosis test datasets, respectively. No significant decrease in accuracy was observed (p > 0.10) when a small training sample size was used. The cyclic training of the discriminator was reduced with self-supervised rotation loss (p < 0.01). The proposed approach is promising for segmenting multiple lung lesion types in CT images.
提出一种针对CT图像中肺部病变（包括肺结节、肺炎和肺结核）的通用分割方法将提高放射学的效率。然而，生成对抗网络的性能受到注释样本可用性有限以及判别器灾难性遗忘的阻碍，而传统的基于形态学的方法的通用性不足以分割各种肺部病变。设计了一种带有上下文感知金字塔特征提取模块的级联双注意力网络来应对这些挑战。设计了一种自监督旋转损失来减轻判别器遗忘。所提出的模型在多中心肺炎、肺结节和肺结核测试数据集上分别实现了70.92%、73.55%和68.52%的骰子系数。当使用小训练样本量时，未观察到准确性有显著下降（p>0.10）。通过自监督旋转损失减少了判别器的循环训练（p<0.01）。所提出的方法在分割CT图像中的多种肺部病变类型方面很有前景。
