Describe how to segment a bright object on a dark background using active contours? How will the contour be represented? How is the topology maintained (to avoid the line to cross itself)?

Given a set of images of some organ (e.g. kidney), how do you build a shape model for the boundary? How is the dimension reduced? How can such a model be used for segmentation?

What descriptors would you use to distinguish different textures in an ultrasound image?

How are the wavelet texture descriptors calculated?

Define the Frangi vesselness descriptor, explain the different terms and how they are calculated.

Describe a deep learning algorithm for detection of lung nodules. How is the performance of such an algorithm evaluated?

Describe the “demons” algorithm for image registration for nonlinear image registration. How can the smoothness of the deformation be controlled.

What are the input and output of the neural network in the Naylor's nuclei segmentation method. What is the reason of this choice? What other formulations are used?