Landmarks (Size: 0b)
> In community MIDAS/Insight Toolkit/Development/Medical/4D CT - Liver - with segmentations

Manually defined landmarks were made on the '0' and '50' images for 4/5 patients. There are approximately 75 manually identified landmarks for each image: approximately 55 on the vessel/airway bifurcations in the lungs, and approximately 20 on uniquely identifiable points inside the abdomen or heart.

Landmarks were manually identified by an engineer (Danielle Pace, Kitware Inc.) using 3D Slicer. All landmarks were verified on all three slice views.

The data also includes preprocessed images, plus the intermediate results of the preprocessing. (1) Crop - crop data to leave smaller border around chest/abdomen (to speed registration) (2) Threshold - to -980,215 (3) Normalize - to mean 0 and variance 1 (4) Scaled - scale to [0,1] using the minimum and maximum intensity from previous 'normalize' step (5) Linear - linear resampling to isotropic spacing

Very important: When you load the landmarks and preprocessed images into 3D Slicer, you must select 'Centered' in the options on the right hand side of the file dialog. Otherwise the landmarks will not line up with the image. If you calculate Target Registration Error, you also must compensate for the image origin yourself.

If you use these landmarks, please cite the following paper: Danielle F. Pace, Stephen R. Aylward, Marc Neithammer, A locally adaptive regularization based on anisotropic diffusion for deformable image registration of sliding organs. IEEE Transactions on Medical Imaging, 32(11), pp. 2114-2126, 2013.


1 community within this community

Empty collection (0b)