Synthetic Aperture Radar (SAR) systems perform signal acquisition under varying incidence angles and register an implicit intensity decay from near to far range. Due to the geometrical interaction between microwaves and the imaged targets, the rates at which intensities decay depend on the nature of the targets, thus rendering single-rate image correction approaches only partially successful. The decay, also known as the incidence angle effect, impacts the segmentation of wide-swath images performed on absolute intensity values. We propose to integrate the target-specific intensity decay rates into a non-stationary statistical model, for use in a fully automatic and unsupervised segmentation algorithm. We demonstrate this concept by assuming Gaussian distributed log-intensities and linear decay rates, a fitting approximation for the smooth, systematic decay observed for extended flat targets. The segmentation is performed on Sentinel-1, Radarsat-2 and UAVSAR wide-swath scenes containing open water, sea ice and oil slicks. As a result, we obtain segments connected throughout the entire incidence angle range, thus overcoming the limitations of modeling that does not account for different per-target decays. The model simplicity also allows for short execution times and presents the segmentation approach as a potential operational algorithm. In addition, we estimate the log-linear decay rates and examine their potential for a physical interpretation of the segments.

Purpose: The registration of a 3D atlas image to 2D radiographs enables 3D pre-operative planning without the need to acquire costly and high-dose CT-scans. Recently, many deep-learning-based 2D/3D registration methods have been proposed which tackle the problem as a reconstruction by regressing the 3D image immediately from the radiographs, rather than registering an atlas image. Consequently, they are less constrained against unfeasible reconstructions and have no possibility to warp auxiliary data. Finally, they are, by construction, limited to orthogonal projections. Methods: We propose a novel end-to-end trainable 2D/3D registration network that regresses a dense deformation field that warps an atlas image such that the forward projection of the warped atlas matches the input 2D radiographs. We effectively take the projection matrix into account in the regression problem by integrating a projective and inverse projective spatial transform layer into the network. Results: Comprehensive experiments conducted on simulated DRRs from patient CT images demonstrate the efficacy of the network. Our network yields an average Dice score of 0.94 and an average symmetric surface distance of 0.84 mm on our test dataset. It has experimentally been determined that projection geometries with 80 $$^{\circ }$$∘ to 100 $$^{\circ }$$∘ projection angle difference result in the highest accuracy. Conclusion: Our network is able to accurately reconstruct patient-specific CT-images from a pair of near-orthogonal calibrated radiographs by regressing a deformation field that warps an atlas image or any other auxiliary data. Our method is not constrained to orthogonal projections, increasing its applicability in medical practices. It remains a future task to extend the network for uncalibrated radiographs.

Deep learning-based (DL) solutions are increasingly been adopted for 2D/3D registration as they can achieve faster 3D reconstructions from 2D radiographs compared to classical methods. This study proposes a novel semi-supervised DL-network for 2D/3D registration, in which an atlas is registered to two orthogonal radiographs. The deformation of the atlas is composed of an affine transformation and a local deformation constrained by a B-spline-based statistical deformation model. The network has been validated on digitally reconstructed radiographs of femur CT images.

Convolutional neural networks (CNN) are popular for segmentation and classification of bones in radiology. However, their training typically requires a database of thousands of manually segmented experimental images. In many cases, such a large dataset is not readily available in the community. In addition, manual segmentation is often too time intensive and prone to human perception, especially in cases of low image quality. In this paper, we show that a CNN can be accurately trained on the digitally reconstructed radiographs (DRR) of a 3D articulating shape model of the object of interest, bypassing the need for a manually-segmented database. The articulating model ensures a realistic appearance of the bones of interest in the DRR, thereby providing suitable training data for segmentation. As a proof-of-concept, we train a CNN on DRRs with the purpose of segmenting the phalanges of a horse leg from radiographs and show that it outperforms a geodesic active contour segmentation method in this particular case. Our proposed training procedure is effective for articulating objects and the resulting CNN can then be applied to real-data segmentation tasks, if preceded by appropriate augmentation.

Most digital models of the equine distal limb that are available in the community are static and/or subject specific; hence, they have limited applications in veterinary research. In this paper, we present an articulatable model of the entire equine distal limb based on statistical shape modeling. The model describes the inter-subject variability in bone geometry while maintaining proper jointspace distances to support model articulation toward different poses. Shape variation modes are explained in terms of common biometrics in order to ease model interpretation from a veterinary point of view. The model is publicly available through a graphical user interface (https://github.com/jvhoutte/equisim) in order to facilitate future digitalization in veterinary research, such as computer-aided designs, three-dimensional printing of bone implants, bone fracture risk assessment through finite element methods, and data registration and segmentation problems for clinical practices.