基于预训练卷积神经网络的迁移学习对原发性肝癌的动态对比增强 CT 诊断：多期图像配准是否必要？
Dynamic contrast-enhanced computed tomography diagnosis of primary liver cancers using transfer learning of pretrained convolutional neural networks: Is registration of multiphasic images necessary?
机构信息
Department of Radiology, Shinshu University School of Medicine, 3-1-1 Asahi, Matsumoto, Nagano, 390-8621, Japan.
出版信息
Int J Comput Assist Radiol Surg. 2019 Aug;14(8):1295-1301. doi: 10.1007/s11548-019-01987-1. Epub 2019 May 3.
PURPOSE
To evaluate the effect of image registration on the diagnostic performance of transfer learning (TL) using pretrained convolutional neural networks (CNNs) and three-phasic dynamic contrast-enhanced computed tomography (DCE-CT) for primary liver cancers.
METHODS
We retrospectively evaluated 215 consecutive patients with histologically proven primary liver cancers, including six early, 58 well-differentiated, 109 moderately differentiated, 29 poorly differentiated hepatocellular carcinomas (HCCs), and 13 non-HCC malignant lesions containing cholangiocellular components. We performed TL using various pretrained CNNs and preoperative three-phasic DCE-CT images. Three-phasic DCE-CT images were manually registered to correct respiratory motion. The registered DCE-CT images were then assigned to the three color channels of an input image for TL: pre-contrast, early phase, and delayed phase images for the blue, red, and green channels, respectively. To evaluate the effects of image registration, the registered input image was intentionally misaligned in the three color channels by pixel shifts, rotations, and skews with various degrees. The diagnostic performances (DP) of the pretrained CNNs after TL in the test set were compared by three general radiologists (GRs) and two experienced abdominal radiologists (ARs). The effects of misalignment in the input image and the type of pretrained CNN on the DP were statistically evaluated.
RESULTS
The mean DPs for histological subtype classification and differentiation in primary malignant liver tumors on DCE-CT for GR and AR were 39.1%, and 47.9%, respectively. The highest mean DPs for CNNs after TL with pixel shifts, rotations, and skew misalignments were 44.1%, 44.2%, and 43.7%, respectively. Two-way analysis of variance revealed that the DP is significantly affected by the type of pretrained CNN (P = 0.0001), but not by misalignments in input images other than skew deformations.
CONCLUSION
TL using pretrained CNNs is robust against misregistration of multiphasic images and comparable to experienced ARs in classifying primary liver cancers using three-phasic DCE-CT.
目的
评估图像配准对使用预训练卷积神经网络（CNN）和三相动态对比增强 CT（DCE-CT）进行原发性肝癌转移学习（TL）的诊断性能的影响。
方法
我们回顾性评估了 215 例经组织学证实的原发性肝癌患者，包括 6 例早期、58 例分化良好、109 例中度分化、29 例低分化肝细胞癌（HCC）和 13 例包含胆管细胞成分的非 HCC 恶性病变。我们使用各种预训练的 CNN 和术前三相 DCE-CT 图像进行 TL。将三相 DCE-CT 图像手动配准以校正呼吸运动。然后，将配准的 DCE-CT 图像分配给 TL 的三个颜色通道：预对比、早期和延迟期图像分别分配给蓝色、红色和绿色通道。为了评估图像配准的效果，将配准的输入图像在三个颜色通道中通过像素移位、旋转和倾斜以不同程度故意错位。三位普通放射科医师（GR）和两位经验丰富的腹部放射科医师（AR）比较了 TL 后预训练 CNN 在测试集中的诊断性能（DP）。统计评估了输入图像的错位和预训练 CNN 的类型对 DP 的影响。
结果
GR 和 AR 对 DCE-CT 原发性恶性肝肿瘤组织学亚型分类和分化的平均 DP 分别为 39.1%和 47.9%。在像素移位、旋转和倾斜配准错位的情况下，TL 后 CNN 的平均 DP 最高分别为 44.1%、44.2%和 43.7%。双向方差分析显示 DP 显著受预训练 CNN 类型的影响（P=0.0001），但不受除斜变形以外的输入图像配准的影响。
结论
使用预训练 CNN 的 TL 对多相图像的配准错误具有鲁棒性，在使用三相 DCE-CT 对原发性肝癌进行分类方面与经验丰富的 AR 相当。
Zotero 插件