We exploit machine-learning formalism to make advance current state-of-the-art methods for dedicated image-registration and motion-estimation tasks. In this context, we are tackling the following challenges:
- exploiting geometrical prior such spherical registration networks for organs with convoluted geometry as dedicated regularization;
- developping dedicated deep learning motion estimator trained with realistic synthetic data generated from a physical simulator to integrate intrinsic properties of the studied modality;
- developping pathological-data generation methods to integrate functional and physiological prior knowledge of the considered pathologies.