New release of the Deep Learning Toolkit: simple AI for biomedical images analysis

TLDR: The Deep Learning Tool Kit (DLTK) for Medical Imaging comes back with a new, simpler version and a model zoo!

DLTK is a neural networks toolkit written in python, on top of TensorFlow.
It is developed to enable fast prototyping with a low entry threshold and ensure reproducibility
in image analysis applications, with a particular focus on medical imaging.

Its goal is to provide the community with state of the art methods and models and to accelerate research in this exciting field.

Examples of use of DLTK for medical applications are:

• Age regression
• Gender classification
• Organ Segmentation
• Image Super-resolution
• as well as a simple GANs.

Two different views of an automatically produced abdominal segmentation (contours each organ in the abdomen and identifies it with a specific label/colour) produced by the toolkit:

a closer look at the abdominal segmentation from a CT scan:

DLTK has performed well in multiple AI competitions (think Kaggle but for academics):

• it tops the free competition in this abdominal segmentation challenge
• was part of the winning entry from the BioMedIA group in the
  BRATS challenge.

DLTK v0.2 brings several major improvements over the previous release, including:

• A medical model zoo, with (re-)implementations of published work and downloadable pre-trained models
• New, full application examples providing a low entry threshold to deep learning methods on medical images (i.e. regression, image super-resolution, etc.)
• Pre-built, flexible state-of-the-art network implementations (e.g. U-Net, ResNet, etc.)
• Simplified interfacing with medical imaging formats (Nifti, ...)
• a full integration with TensorFlow Estimators

The complete list of changes is available here: https://github.com/DLTK/DLTK/blob/master/CHANGELOG.md

source: https://github.com/DLTK/DLTK
Pypi: https://pypi.python.org/pypi/dltk
twitter: https://twitter.com/dltk_
gitter: https://gitter.im/DLTK/DLTK
reddit:
License: Apache v2.0
paper: https://arxiv.org/abs/1711.06853v1

Note: content reused with the permission of the authors