Fast Lesion Extraction using Convolutional Neural Networks

FLEXCONN (Fast Lesion Extraction using Convolutional Neural Networks) is a toolbox for segmenting white matter lesions from multi-contrast MR images. Using T1-w and FLAIR images, a fully convolutional neural network (CNN) is trained using manually labeled training data. The trained CNN model can be applied to pre-processed pair of T1 and FLAIR images to generate a lesion membership as well as a hard segmentation.

Python codes for training and prediction are provided. Trained models using manually labeled atlases from ISBI-2015 challenge (https://www.nitrc.org/projects/longitudi...) are also provided. The CNN is implemented in Tensorflow and Keras (https://github.com/fchollet/keras).
Download Now Download Now
OR See All Files

Specifications

License:Academic Free License ("AFL")
Domain:MR

Recent Activity

Show: 
File Activity

flexconn: Code and Examples release

FLEXCONN1.1.zip posted by Snehashis Roy on Oct 23

File Activity

flexconn: Code and Examples release

FLEXCONN1.0.zip posted by Snehashis Roy on Oct 18

File Activity

flexconn: Atlases release

ISBI2015_Challenge_atlases.zip posted by Snehashis Roy on Oct 16