Four-dimensional image registration for image-guided radiotherapy

Newly emerged four-dimensional (4D) imaging techniques such as 4D-computed tomography (CT), 4D-cone beam CT, 4D-magnetic resonance imaging, and 4D-positron emission tomography are effective tools to reveal the spatiotemporal details of patients' anatomy. To use the 4D data acquired under different conditions or using different modalities, an algorithm for registering 4D images must be in place. We developed an automated 4D-4D registration method to take advantage of 4D information.

We used 4D-4D matching to find the appropriate three-dimensional anatomy in the fixed image for each phase of the moving image and spatially register them. A search algorithm was implemented to simultaneously find the best phase and spatial match of two 4D inputs. An interpolation scheme capable of deriving an image set based on temporally adjacent three-dimensional data sets was developed to deal with the situation in which the discrete temporal points of the two inputs do not coincide or correspond.

In a phantom study, our technique was able to reproduce the known "ground truth" with high spatial fidelity. The technique regenerated all deliberately introduced "missing" three-dimensional images at different phases of the input using temporal interpolation. In the registration of gated-magnetic resonance imaging and 4D-CT, the algorithm was able to select the appropriate CT phase. The technique was also able to register 4D-CT with 4D-cone beam CT and two 4D-CT scans acquired at different times. A spatial accuracy of <3 mm was achieved in 98% of voxels in all cases.

Automated 4D-4D registration can find the best possible spatiotemporal match between two 4D data sets and is useful for image-guided radiotherapy applications.