General information
Medical Image Computing Coursework 2016
For the coursework, you are given two IPython notebook files that contain descriptions and some skeleton code. You will need to add your implementations at the specified locations directly within the notebooks. Data for testing your implementations is provided.
SUBMISSION
You will need to upload an archive (.zip or .tar.gz) with the name ¡®coursework¡¯ to CATE. The archive should only contain the two notebook files ¡®part1.ipynb¡¯ and ¡®part2.ipynb¡¯. To minimise the size of the files, you can clear all cell outputs before compression by choosing from the menu ¡®Cell -> All Output -> Clear¡¯.
Submission deadline on CATE: Friday 02 Dec 2016 23:59 Age regression from brain MRI
The objective for the coursework is to implement two different approaches for age regression from brain MRI data. Data from 300 subjects will be provided. For each approach, you are given a dedicated IPython notebook in which you are asked to complete five tasks that are highlighted in the notebook files.
PART I
The first approach will make use of gray matter masks that have been extracted from a set of MRI scans and registered to a reference space for spatial normalisation. Each voxel in the gray matter mask is a feature. The masks will be provided. Because those masks are quite large, a dimensionality reduction using PCA should be performed first before training a suitable regressor. The performance is evaluated via cross validation.
PART II
The second approach aims to regress the age of a subject from the volume of the lateral ventricles. To this end, a ventricle segmentation needs to be implemented that allows extraction of an image mask that contains ventricle voxels from which the volume in millilitres can be calculated. But before segmentation, the brain images need to be registered to an atlas for which a registration method needs to be implemented. Finally, a suitable regressor needs to be trained that allows to predict age from ventricle volume. The performance is evaluated via cross validation.
Useful resources
Dataset: http://www.doc.ic.ac.uk/~bglocker/teaching/mic-coursework-data.zip
PCA: http://scikit-learn.org/stable/modules/generated/sklearn.decomposition.PCA.html Regression: http://scikit-learn.org/stable/supervised_learning.html
Segmentation: http://goo.gl/W6EO5u and https://sites.google.com/site/mrilateralventricle/