Ledig et al.

From CADDementia
Jump to: navigation, search

Dementia classification of AD, MCI and controls for the CADDementia dataset

C. Ledig1, R. Guerrero1, T. Tong1, K.R. Gray1, A. Schmidt-Richberg1, A. Makropoulos1,2, R.A. Heckemann3,4, D. Rueckert1

1 Department of Computing, Imperial College London, UK

2 Division of Imaging Sciences and Biomedical Engineering, King’s College London, UK

3 MedTech West, Institute of Neuroscience and Physiology, University of Gothenburg, Sweden

4 Division of Brain Sciences, Faculty of Medicine, Imperial College London, UK



1. Image pre-processing

2. Structural Segmentation / Volumetric Analysis

3. Cortical Features

4. Grading Features

5. Intensity Features

6. Classification experiments

Explanation: A random forest classifier [Breiman, 2001] is trained on the 734 ADNI subjects for which all features are available. The trained classifier is then applied to the CADDementia test subjects. This method is able to predict either binary labels or class probabilities, and can be directly applied for three-class classification.

Software: scikit-learn [Pedregosa et al., 2011]


Initialise random forest classifier with the following parameters (all others take default values):

  • 100 trees
  • entropy splitting criterion
               rf = RandomForestClassifier(n_estimators = 100, criterion='entropy')

Train classifier using ADNI data and labels:

               rf = rf.fit(ADNIdata,ADNIlabels)

Apply trained classifier to CADD test data:

               binout = rf.predict(CADDdata)
               probout = rf.predict_proba(CADDdata)




  • L. Breiman, Random Forests, Machine Learning, vol.45, no.1, pp.5–32, 2001.
  • F. Pedregosa, G. Varoquaux, A. Gramford, et al., Scikit-learn: Machine Learning in Python, The Journal of Machine Learning Research, vol.12, pp.2825–2830, 2011.