i) To utilize tools of geometry and topology in the bone quality assessment;

ii) to define new micro-scale fragility index as well as to detect cracks by using topological and geometrical methods. Those techniques will be developed in parallel and compared with AI strategies to obtain the best possible performance;

iii) to determine the best filtration of the input image to obtain the cleanest discrimination of patients and control group;

iv) to apply the tools based on relative persistent homology, or max-flow-min-cut duality to detect cracks;

v) to adopt Morse-Smale function of a merge tree of an appropriate filtration for the learning tasks.

One of the main targets (in the framework of WP3) is to define new bone descriptors that can be computed given a gray-scale image of the bone, obtained from a synchrotron, or more standard technique like micro-CT or similar.

Enrolment: IMPAN

Secondments:

1) Prof. Vergani (POLIMI): Analyses of bone images

2) Prof. Schwiedrzik (EMPA): experimental test to measure micro-structural parameters of bones. 

i) a scalable, well documented and efficient implementation that take such an image at the input;

ii) to estimate the bone quality index using methods of computational geometry and topology. Those methods should be stable with respect to noise that may be present due to inaccuracy of image acquisition techniques;

iii) to provide a fully automated algorithm to detect cracks in bone and to assess the severity of the cracks. The input is a gray-scale image of a bone while the output is an annotation of the image, that indicate the location of cracks. The algorithm should allow tracking the evolution of a single cracks across images, and provide an automatic tool to assess the severity of the cracks.