Pose-dependent weights and Domain Randomization for fully automatic X-ray to CT Registration

Fully automatic X-ray to CT registration requires a solid initialization to provide an initial alignment within the capture range of existing intensity-based registrations. This work adresses that need by providing a novel automatic initialization, which enables end to end registration. First, a neural network is trained once to detect a set of anatomical landmarks on simulated X-rays. A domain randomization scheme is proposed to enable the network to overcome the challenge of being trained purely on simulated data and run inference on real Xrays. Then, for each patient CT, a patient-specific landmark extraction scheme is used. It is based on backprojecting and clustering the previously trained networks predictions on a set of simulated X-rays. Next, the network is retrained to detect the new landmarks. Finally the combination of network and 3D landmark locations is used to compute the initialization using a perspective-n-point algorithm. During the computation of the pose, a weighting scheme is introduced to incorporate the confidence of the network in detecting the landmarks. The algorithm is evaluated on the pelvis using both real and simulated x-rays. The mean (+-standard deviation) target registration error in millimetres is 4.1 +- 4.3 for simulated X-rays with a success rate of 92% and 4.2 +- 3.9 for real X-rays with a success rate of 86.8%, where a success is defined as a translation error of less than 30mm.

[1]  Keiichi Abe,et al.  Topological structural analysis of digitized binary images by border following , 1985, Comput. Vis. Graph. Image Process..

[2]  M. Powell The BOBYQA algorithm for bound constrained optimization without derivatives , 2009 .

[3]  Cong Gao,et al.  Automatic annotation of hip anatomy in fluoroscopy for robust and efficient 2D/3D registration , 2019, International Journal of Computer Assisted Radiology and Surgery.

[4]  Rui Liao,et al.  Unsupervised X-ray image segmentation with task driven generative adversarial networks , 2020, Medical Image Anal..

[5]  Wojciech Zaremba,et al.  Domain randomization for transferring deep neural networks from simulation to the real world , 2017, 2017 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS).

[6]  Varun Ramakrishna,et al.  Convolutional Pose Machines , 2016, 2016 IEEE Conference on Computer Vision and Pattern Recognition (CVPR).

[7]  Jia Deng,et al.  Stacked Hourglass Networks for Human Pose Estimation , 2016, ECCV.

[8]  M. Viergever,et al.  Robust initialization of 2D-3D image registration using the projection-slice theorem and phase correlation. , 2010, Medical physics.

[9]  Z. Jane Wang,et al.  Real-time 6DoF pose recovery from X-ray images using library-based DRR and hybrid optimization , 2016, International Journal of Computer Assisted Radiology and Surgery.

[10]  Nassir Navab,et al.  Initialize globally before acting locally: Enabling Landmark-free 3D US to MRI Registration , 2018, MICCAI.

[11]  Bostjan Likar,et al.  A review of 3D/2D registration methods for image-guided interventions , 2012, Medical Image Anal..

[12]  Mathias Unberath,et al.  Towards Fully Automatic X-Ray to CT Registration , 2019, MICCAI.

[13]  Nassir Navab,et al.  Enabling machine learning in X-ray-based procedures via realistic simulation of image formation , 2019, International Journal of Computer Assisted Radiology and Surgery.

[14]  Amir Alansary,et al.  Predicting Slice-to-Volume Transformation in Presence of Arbitrary Subject Motion , 2017, MICCAI.

[15]  Cheng Chen,et al.  PnP-AdaNet: Plug-and-Play Adversarial Domain Adaptation Network at Unpaired Cross-Modality Cardiac Segmentation , 2019, IEEE Access.

[16]  Nassir Navab,et al.  DeepDRR - A Catalyst for Machine Learning in Fluoroscopy-guided Procedures , 2018, MICCAI.

[17]  Nassir Navab,et al.  X-ray-transform Invariant Anatomical Landmark Detection for Pelvic Trauma Surgery , 2018, MICCAI.

[18]  Rui Liao,et al.  Dilated FCN for Multi-Agent 2D/3D Medical Image Registration , 2017, AAAI.