Transfer learning from an artificial radiograph-landmark dataset for registration of the anatomic skull model to dual fluoroscopic X-ray images.
Transfer learning from an artificial radiograph-landmark dataset for registration of the anatomic skull model to dual fluoroscopic X-ray images.
复制标题
DOI：
10.1016/j.compbiomed.2021.104923
复制
发表时间：
2021-11
影响因子：
7.7
通讯作者：
Li G
中科院分区：
文献类型：
--
作者：
Zhou C;Cha T;Peng Y;Li G
将3D解剖结构配准到其2D双荧光透视X射线图像是广泛使用的运动跟踪技术。然而，深度学习的实现往往受到缺乏医学图像和地面实况的阻碍。在这项研究中，我们提出了一种使用从人工数据集训练的深度神经网络进行3D到2D配准的迁移学习策略。根据女性受试者的颅脑CT数据自动创建数字重建X线片（DRR）和放射学颅骨标志。它们被用来训练用于地标检测的残差网络（ResNet）和周期生成对抗网络（GAN），以消除DRR和实际X射线之间的风格差异。ResNet检测到经历GAN风格平移的X射线上的地标，并将其用于实际双荧光镜图像中颅骨3D到2D配准的三角测量优化（具有非正交设置、点X射线源、图像失真和部分捕获的颅骨区域）。在颅颈运动的多个场景中评价配准精度。在行走中，基于学习的颅骨配准的角度/位置误差为3.9±2.1°/4.6±2.2 mm。然而，由于在末端位置的双透视图像上成像的颅骨区域过小，因此在功能性颈部活动期间的准确度较低。策略性地增强人工训练数据的方法可以解决复杂的颅骨配准场景，并且有潜力扩展到广泛的配准场景。
Registration of 3D anatomic structures to their 2D dual fluoroscopic X-ray images is a widely used motion tracking technique. However, deep learning implementation is often impeded by a paucity of medical images and ground truths. In this study, we proposed a transfer learning strategy for 3D-to-2D registration using deep neural networks trained from an artificial dataset. Digitally reconstructed radiographs (DRRs) and radiographic skull landmarks were automatically created from craniocervical CT data of a female subject. They were used to train a residual network (ResNet) for landmark detection and a cycle generative adversarial network (GAN) to eliminate the style difference between DRRs and actual X-rays. Landmarks on the X-rays experiencing GAN style translation were detected by the ResNet, and were used in triangulation optimization for 3D-to-2D registration of the skull in actual dual-fluoroscope images (with a non-orthogonal setup, point X-ray sources, image distortions, and partially captured skull regions). The registration accuracy was evaluated in multiple scenarios of craniocervical motions. In walking, learning-based registration for the skull had angular/position errors of 3.9±2.1°/4.6±2.2 mm. However, the accuracy was lower during functional neck activity, due to overly small skull regions imaged on the dual fluoroscopic images at end-range positions. The methodology to strategically augment artificial training data can tackle the complicated skull registration scenario, and has potentials to extend to widespread registration scenarios.
登录
查看更多内容
影响因子：
2.4
作者：
Englander ZA;Martin JT;Ganapathy PK;Garrett WE;DeFrate LE
通讯作者：
DeFrate LE
强大的3D-2D图像注册：在解剖变形的存在下应用于脊柱干预和椎骨标记。
影响因子：
3.5
作者：
Otake Y;Wang AS;Webster Stayman J;Uneri A;Kleinszig G;Vogt S;Khanna AJ;Gokaslan ZL;Siewerdsen JH
通讯作者：
Siewerdsen JH
DOI：
10.1115/1.2205865
发表时间：
2006-08-01
影响因子：
1.7
作者：
Bingham, Jeffrey;Li, Guoan
通讯作者：
Li, Guoan
影响因子：
3.8
作者：
Tsai, Tsung-Yuan;Lu, Tung-Wu;Hsu, Horng-Chaung
通讯作者：
Hsu, Horng-Chaung
DOI：
10.1115/1.4036311
发表时间：
2017-06-01
影响因子：
1.7
作者：
Yu, Yan;Mao, Haiqing;Cha, Thomas D.
通讯作者：
Cha, Thomas D.