Hubless keypoint-based 3D deformable groupwise registration

  title={Hubless keypoint-based 3D deformable groupwise registration},
  author={R{\'e}mi Agier and S{\'e}bastien Valette and Razmig K{\'e}chichian and Laurent Fanton and R{\'e}my Prost},
  journal={Medical image analysis},
Learning 3D medical image keypoint descriptors with the triplet loss
A 3D keypoint descriptor is proposed which is used to match keypoints extracted from full-body CT scans and shows improvement compared to a hand-crafted descriptor as literature has shown that jointly learning a detector and a descriptor gives further performance boost.
Learning 3D Medical Image Patch Descriptors with the Triplet Loss
The proposed methodology is to generate semi-synthetic data by transforming real volumes and to use a triplet loss inspired by 2D descriptor learning, which outperforms the hand-crafted descriptor 3D-SURF, a 3D extension of SURF, with similar runtime.
Local Surf-Based Keypoint Transfer Segmentation
The approach is based on 3D SURF keypoint extraction, instead of 3D SIFT in the original algorithm, which yields a significantly higher number of keypoints, and the resulting segmentation accuracy is significantly increased, and smaller organs can be segmented correctly.
Atlas construction and spatial normalisation to facilitate radiation-induced late effects research in childhood cancer
The findings indicate satisfactory mapping between a heterogeneous group of patients and the template CT, and the poorest performance was for organs in the abdominal and pelvic region, likely due to respiratory and physiological motion and to the highly deformable nature of abdominal organs.
Disentangled representations: towards interpretation of sex determination from hip bone
A new paradigm for better interpretability of class differences was proposed and the features encoded by the model, that distinguish the different classes were found to be consistent with expert knowledge.


Volumetric Image Registration From Invariant Keypoints
The method extends the scale invariant feature transform (SIFT) to arbitrary dimensions by making key modifications to orientation assignment and gradient histograms and rotation invariance is proven mathematically.
Keypoint Transfer for Fast Whole-Body Segmentation
This work introduces an approach for image segmentation based on sparse correspondences between keypoints in testing and training images, which offers a speed-up of about three orders of magnitude in comparison with common multi-atlas segmentation while achieving an accuracy that compares favorably.
Automatic Construction of Parts+Geometry Models for Initializing Groupwise Registration
It is shown that both the model and its matches can be automatically obtained, and that the matches are able to effectively initialize a groupwise nonrigid registration algorithm, leading to accurate dense correspondences.
Feature‐based groupwise registration by hierarchical anatomical correspondence detection
This article proposes a novel feature‐based groupwise registration algorithm to establish the anatomical correspondence across subjects by using the attribute vector that is defined as the morphological signature for each voxel, and achieves more robust and accurate registration results.
Group-Wise Registration of Point Sets for Statistical Shape Models
A novel, fast, group-wise registration technique based on establishing soft correspondences between groups of point sets, capable of learning the shape variations within a training set, and compared to available state-of-the-art group- wise registration algorithms including feature-based and volume-based approaches.
Spline-Based Image Registration
A new registration algorithm based on spline representations of the displacement field which can be specialized to solve all of the problems in multiframe image analysis, including the computation of optic flow, stereo correspondence, structure from motion, and feature tracking.
Hubless 3D Medical Image Bundle Registration
This work proposes a hubless medical image registration scheme, able to conjointly register massive amounts of images, and provides an eye-detection application as a first step to patient image anonymization.
A new point matching algorithm for non-rigid registration