Department of Ophthalmology, Shanghai General Hospital, National Clinical Research Center for Eye Diseases, Shanghai Jiao Tong University School of Medicine, Shanghai Key Laboratory of Ocular Fundus Disease, Shanghai Engineering Center for Visual Science and Photomedicine, Shanghai, 200080, China.
Department of Ultrasound, Shanghai Pulmonary Hospital, Tongji University School of Medicine, Shanghai, 200433, China.
J Digit Imaging. 2020 Oct;33(5):1155-1166. doi: 10.1007/s10278-020-00356-8.
To evaluate the application of machine learning for the detection of subpleural pulmonary lesions (SPLs) in ultrasound (US) scans, we propose a novel boundary-restored network (BRN) for automated SPL segmentation to avoid issues associated with manual SPL segmentation (subjectivity, manual segmentation errors, and high time consumption). In total, 1612 ultrasound slices from 255 patients in which SPLs were visually present were exported. The segmentation performance of the neural network based on the Dice similarity coefficient (DSC), Matthews correlation coefficient (MCC), Jaccard similarity metric (Jaccard), Average Symmetric Surface Distance (ASSD), and Maximum symmetric surface distance (MSSD) was assessed. Our dual-stage boundary-restored network (BRN) outperformed existing segmentation methods (U-Net and a fully convolutional network (FCN)) for the segmentation accuracy parameters including DSC (83.45 ± 16.60%), MCC (0.8330 ± 0.1626), Jaccard (0.7391 ± 0.1770), ASSD (5.68 ± 2.70 mm), and MSSD (15.61 ± 6.07 mm). It also outperformed the original BRN in terms of the DSC by almost 5%. Our results suggest that deep learning algorithms aid fully automated SPL segmentation in patients with SPLs. Further improvement of this technology might improve the specificity of lung cancer screening efforts and could lead to new applications of lung US imaging.
为了评估机器学习在超声（US）扫描中检测胸膜下肺病变（SPL）的应用，我们提出了一种新的边界恢复网络（BRN），用于自动 SPL 分割，以避免与手动 SPL 分割相关的问题（主观性、手动分割错误和高时间消耗）。共导出了 255 名患者的 1612 个超声切片，这些患者的超声切片中可见 SPL。基于 Dice 相似系数（DSC）、马修斯相关系数（MCC）、Jaccard 相似度量（Jaccard）、平均对称表面距离（ASSD）和最大对称表面距离（MSSD）评估了神经网络的分割性能。我们的双阶段边界恢复网络（BRN）在分割准确性参数方面优于现有的分割方法（U-Net 和全卷积网络（FCN）），包括 DSC（83.45±16.60%）、MCC（0.8330±0.1626）、Jaccard（0.7391±0.1770）、ASSD（5.68±2.70 mm）和 MSSD（15.61±6.07 mm）。在 DSC 方面，它也比原始 BRN 高出近 5%。我们的研究结果表明，深度学习算法有助于对有 SPL 的患者进行全自动 SPL 分割。这项技术的进一步改进可能会提高肺癌筛查工作的特异性，并可能为肺部 US 成像带来新的应用。
