Image registration: models and methods

Moroni D., Salvetti O.

I.4 Image processing and computer vision  J.3 Life and medical sciences  MedicImage Registration  Non-rigid matchingal  Imaging 

This work has been carried on within the project SIMCAR ``A methodology based on multi-source signals and images for the identification of a model oriented toward cardiac surgical simulation.'' The project aims at the identification, development and test of a methodology for the assessment of morpho-functional cardiac parameters, evaluated by means of various diagnostic resources and embedded in a mathematical model oriented toward cardiac surgical simulation. In particular, current medical imaging techniques (such as magnetic resonance imaging and helical CT) provide temporal sequences of morphological data and, thus, information about heart kinematical behavior. At the same time, imaging modalities supply punctual data about tissues characteristic. Other techniques (such as ECG and Pressure/Volume loops) provide a natural temporal reference system as well as global constraints on morpho-functional parameters. Since comparison of the available heterogeneous data is required, the project plan considers a first step focused on analysis and implementation of spatio-temporal registration techniques. This report offers a review, as far as possible exhaustive, of registration methods and models, giving particular prominence to topics, such as non-rigid registration and analysis of images sequences, whose importance is unquestionable in many applications (including cardiac imaging area). Methods are introduced from a general viewpoint that, as we hope, clarifies, a part from limits and values, the basic assumptions of each one. Of course, the set of proposed techniques is finally applied in cardiac image analysis, although some, particulary promising, methods, due to very recent introduction, still lack cardiac applications.

Source: ISTI Technical reports, 2005