Department of Diagnostic Radiology, Tokyo Medical and Dental University, Tokyo, Japan.
Department of Radiology, Dokkyo Medical University, Tochigi, Japan.
J Ultrasound Med. 2021 Jan;40(1):61-69. doi: 10.1002/jum.15376. Epub 2020 Jun 27.
We sought to generate realistic synthetic breast ultrasound images and express virtual interpolation images of tumors using a deep convolutional generative adversarial network (DCGAN).
After retrospective selection of breast ultrasound images of 528 benign masses, 529 malignant masses, and 583 normal breasts, 20 synthesized images of each were generated by the DCGAN. Fifteen virtual interpolation images of tumors were generated by changing the value of the input vector. A total of 60 synthesized images and 20 virtual interpolation images were evaluated by 2 readers, who scored them on a 5-point scale (1, very good; to 5, very poor) and then answered whether the synthesized image was benign, malignant, or normal.
The mean score of overall quality for synthesized images was 3.05, and that of the reality of virtual interpolation images was 2.53. The readers classified the generated images with a correct answer rate of 92.5%.
A DCGAN can generate high-quality synthetic breast ultrasound images of each pathologic tissue and has the potential to create realistic virtual interpolation images of tumor development.
我们试图使用深度卷积生成对抗网络（DCGAN）生成逼真的合成乳房超声图像，并表达肿瘤的虚拟插值图像。
回顾性选择 528 个良性肿块、529 个恶性肿块和 583 个正常乳房的超声图像后，通过 DCGAN 生成了每个肿块的 20 张合成图像。通过改变输入向量的值，生成了 15 张肿瘤的虚拟插值图像。由 2 位读者对总共 60 张合成图像和 20 张虚拟插值图像进行评估，读者对每张图像的质量进行 5 分制评分（1 分为非常好，5 分为非常差），然后回答合成图像是良性、恶性还是正常。
合成图像整体质量的平均得分为 3.05，虚拟插值图像的逼真度得分为 2.53。读者对生成的图像的正确回答率为 92.5%。
DCGAN 可以生成高质量的每个病理组织的合成乳房超声图像，并具有生成肿瘤发展的逼真虚拟插值图像的潜力。
