This assignment involves the detection of white matter lesions in the brain from MR images. Three different types of MR images are used as input (T1-weighted, T2-weighted and FLAIR-weighted). See image above for an example (FLAIR-weighted and annotated image, which is the ground truth).

50 2D slices of scans are available (200 images in total). Classifications are made on a pixel level with pixel based features, like the brightness value, the distance transform from the brain edge, a blobness measure and more.

A dice similarity coefficient (DSC) greater than 0.5 was achieved.

• Python 2.7
• scikit-learn ML stack (sklearn, numpy, scipy, matplotlib, skimage)
• OpenCV2
• hickle (pickle or cPickle can be used instead)
• tqdm

• Robbert van der Gugten (@robbertvdg)
• Inez Wijnands (@Moorkopsoesje)
• Guido Zuidhof (@gzuidhof)