TU Wien:Medizinische Bildverarbeitung VO (Langs)/Prüfung 2020-07-29
1.How can you evaluate methods that identify a specific disease based on patient data (i.e. a classifier of patients). List two evaluation measures necessary to asses and compare methods. Briefly describe their relationship.
2. You perform an fMRI experiment with 30 control subjects. For each subject you acquire fMRI data (I_1, …, I_30) and structural MRI data (S_1, …, S_30) of the brain. You want to compare the fMRI signals across subjects. Which registrations between which data do you perform? For each please give (a) name of transformation, (b) voxel similarity measure.
3. Which material properties does CT measure? Which transformation is necessary to reconstruct 3D imaging data from the measured signals?
4. Outline an algorithm (both training and application to new data) for the detection of lesions with the help of a classifier.
5. What is the difference between a General Linear Model and Multi Voxel Pattern Analysis? Give one application for each and describe one of them in more detail.
6. Which appearance information of the training examples is learned by Active Shape Models? How does the search in new data work?
7. How can you identify groups of similar examples in the data set with help of an Auto Encoder?
8. Which modality is used to visualize bundles of nerve fibres?
9. Sketch the two distributions of the point sets A and B having covariance matrices A = [4 0; 0 1] and B = [5 1; 1 2] in a 2D-space. For which does PCA make sense? (i.e. where would PCA actually change the coordinate system?)
10. How can you train a decision tree in a random forest? Please sketch the process. How are the training examples selected that are used to train an individual decision tree? How are the final classification results composed?