Background: Three Point Dixon technique is a methodology able to separate water and fat components within magnetic resonance images, exploiting the phase of the complex data. It works exploiting the phase differences between three MRI images acquired with different Echo Times. Within the procedure, the most problematic step is the so called phase unwrapping operation.
Methods: We propose a new approach for unwrapping the acquired image phases. The presented technique works directly on each single complex image instead of phase differences, allowing reconstructions with high accuracy. The methodology jointly exploits the amplitude and the phase information. For assuring high computational efficiency and fast convergence to the global optimal solution, a graph cuts based optimization approach has been implemented.
Results: The proposed technique has been first applied to a phantom simulating a magnetic resonance image of the human head. Subsequently, the algorithm has been tested on a real data set consisting of three head images acquired in axial position. In both cases, results have been compared with the ones obtained using classical unregularized version of three point Dixon technique.
Conclusions: A novel, computationally fast and effective algorithm for phase unwrapping in three point Dixon water and fat separation is presented. The methodology provides regularized fat and water component estimation with a higher accuracy compared to the classical approach.