Difficult Image Registration Data

NAMIC Project Page

Goal: Identify practical solutions to difficult atlas-to-subject image registration problems.
Task: Register human brain MR and body CT atlases to subject images with maximal accuracy and minimal manual supervision.
Challenges: Unknown inintialization, lack of one-to-one correspondence due to pathology, injury, image truncation, inter-subject and inter-species variation.
Output: Registration solutions, including a description of algorithms and parameters used. Solutions may possibly take a variety of forms, e.g. linear or
non-linear transforms, landmark or feature correspondences, etc. Precise registration may not generally exist throughout the images, however solutions
should align structure present in both images being registered.

MR - Brain

Image Description Data Source
Atlas: human brain average MR. Data OASIS
Subject 1: Traumatic brain injury. Data NAMIC
Subject 2: Intra-operative, unspecified rotation. Data Archip et al.
Subject 3: Rhesus Macaque. Data Montreal Neurological Institute

CT - Body

Image Description Data Source
Atlas: human body CT. Data National Biomedical Imaging Archive
Subject 1: Torso. Data MIDAS Image Collection
Subject 2: Torso, legs. Data Public CT DICOM Collection