Better examples, ANTSpy, microscopy tools

In this release we included a new series of examples using down-scaled data for faster testing of the nighres installation and core algorithms. Most of them should run on a laptop! The old, full resolution examples are still available.

The ANTS utilities in the registration tools now use ANTSpy, which is generally easier to install. The corresponding modules are identical except that 'antsreg' is replaced by 'antspy'.

New features have been added to the registration modules, among others one to automatically select the number of multiscale steps in ANTS based on a desired scaling factor, and one to build smooth transitions of a given thickness at the edge of a mask.

New tools for processing microscopy data have been added, from slice-to-slice intensity normalization to line feature detection and non-local interpolation. Many of these tools work in 2D, 2.5D and 3D.

Subcortical parcellation

• added a new algorithm for subcortical parcellation, MASSP
• improved the robustness of a few methods for statistics (median, percentiles)
• added options for multi-contrast and/or 2D registration with ANTs
• added a new method for (relative) PD mapping with MP2RAGEME
• added example datasets (MP2RAGEME) and atlases (MASSP, AHEAD). New atlases are now by default stored separately and downloaded only if used.

Vessels, denoising, white matter tracts & more

• Added new modules for
  • high-resolution image denoising (Local-Complex PCA; Bazin et al., 2019)
  • vascular modeling (Huck et al., 2019)
  • white matter tract labeling (Bazin et al., 2011)
  • topology correction (Bazin et al., 2007)
  • cortical surface extraction (Han et al., 2003) and inflation (Tosun et al., 2004)
  • tissue classification (Pham et al., 2001).
• Almost all utilities originally in CBSTools have been migrated, unless trivial
• Standard module behaviour is now to either input and output filenames (when saving data) or nifti / mesh objects (when not saving). This optimizes memory usage and simplifies the architecture.
• Rules about filenames have been relaxed.

Fixing pip install issue in release 1.1.0b

Fixes #102
Thanks to @owainkenwayucl for pointing this out!

Python3, IMCN and many new features

This release includes most of the CBSTools package and migrates the code to python 3

New features:

• Python 3 integration: all runs with python 3 now, including JCC (note: because of the different ways python 2 and 3 coexist in different installations (regular Linux, Conda, Docker, etc) you have to be careful to explicitly use 'python3 -m pip', 'python3 -m jcc' etc rather than the shorthand versions. There's a naming convention problem with 'dir' in the init.py as generated by JCC for python 3.7, which we handled with post-hoc find&replace).
• New examples: now including registration and atlas-based segmentation examples. The registration tools require ANTs and Nipype (note: for these modules, the inputs must be file names, not directly Nifti1Image objects).
• New modules: a lot of additional CBSTools have been ported, including a clean port of the isosurface meshing tool. New modules from the IMCN-TK have been added as well (note: because we have multiple sources, now the java code from all external repositories is compiled into a single collection of jar files and wrapped as one "nighresjava" python package).
• Memory management: the JVM checks for available memory and recruits it accordingly.
• File i/o: the .vtk surface files can handle different mesh options (with or without data included, one or n-dimensional).

Docker and bug fixes -- reloaded

Last release was falsely drafted before last commit with version change was applied. New try.

Docker and bug fixes

Merge pull request #79 from juhuntenburg/enh/docs_credit_pics

Enh/docs credit pics

Local build beta release deploy to PyPI

Now PyPI deploy actually works

Local build beta release

1.0.0b5

new beta release

Google summer of code beta release 4 (update NITRC data)

1.0.0b4

new pre-release