In abdominal radiotherapy, respiratory-correlated 4DCT is currently the gold standard for motion controlled imaging of highly mobile tumors. The main drawbacks of such CTs are poor soft tissue contrast and dose burden for the patient. This research aims to develop and validate an accurate 4DMRI method within a clinically relevant acquisition time and T2-weighted contrast for imaging abdominal structures by retrospective sorting of images.

We developed a 4DMRI method by alternating a fast (0.6 seconds per 2D slice) T2-weighted turbo spin echo (TSE) image acquisition (resolution: 1.3 x 1.6 mm²; 5 mm slice thickness) with a 1D navigator acquisition. The navigator obtained the diaphragm position prior to each slice acquisition. During 6 minutes of free breathing, slices were acquired continuously, yielding 60 image frames per slice over a volume of 11 slices.

After the acquisition, each image was coupled to a navigator signal and assigned to a respiratory state by either phase or amplitude binning. The resulting 4DMRI consisted of 110 assigned image states (10 bins, 11 slices). For phase binning, bins were determined by dividing each end-exhale peak to peak position into ten evenly distributed bins (figure 1). For amplitude binning, bins were determined according to the range in diaphragm positions determined by the navigator. The range was defined per volunteer and divided into ten bins. The minima and maxima were the mean diaphragm position at end-inhale and end-exhale, respectively. Data from outside this range was deleted. The two strategies were used to reconstruct 4DMRI images for 10 volunteers (7 female, mean age 28 years) obtained on a 3T scanner.

The position and superior–inferior (SI) motion of the diaphragm were quantified by registering the diaphragm to the begin-inhale image of a series (bin 1). Sorting images into respiratory bins often resulted in multiple images assigned to the same state. From this set, the image with the median diaphragm position was selected for 4DMRI reconstruction and the standard deviations (SD) of positions were calculated. Sometimes, when no images were assigned to a state, an incomplete 4DMRI resulted. The 4DMRIs were evaluated on data completeness (filled states of 4DMRI data set) and intra-bin variation of diaphragm position (mean SD and maximum SD). The variation was calculated over all bins from the three central slices that covered the largest diaphragm motion. The Wilcoxon’s signed rank test was used to test the difference between the two methods.

In order to assess relation between scan time and data completeness, we measured 200 dynamics for a single volunteer and reconstructed multiple 4DMRIs from subsets of these dynamics and evaluated the data completeness per reconstruction.

4DMRI data sets were acquired using a T2-weighted sequence, facilitating abdominal tissue contrast.

Figure 2 shows the relation between the number of dynamics used in the 4DMRI reconstruction versus the data completeness. For this volunteer, data completeness reaches 90% at 60 dynamics for amplitude binning after which the increase of completeness reaches a plateau.

Figure 3 shows a coronal slice of a 4DMRI in one respiratory state corresponding to bin 7. The sagittal reconstruction demonstrates a continuous diaphragm for the amplitude binning method in contrast to a discontinuous shape for the phase binned reconstruction. Generally, the spread on diaphragm position was smaller for the amplitude binning than for the phase binning (difference of 1.6 (3.8) mm for mean (max) SD, both statistically significant with p<0.01) as shown in Table 1. Phase binning resulted in a more complete (4.9%) dataset. For one volunteer for a central slice the diaphragm position variation per respiratory bin is shown in Figure 4, demonstrating the lower intra-bin variation for amplitude binning compared to phase binning. The median image per bin was used for 4DMRI reconstruction and depicts a smooth respiratory signal.

With respect to intra-bin variability, amplitude binning is superior to phase binning. However, phase binning can be acquired in a shorter period given a certain data completeness. For the purpose of accurate tumor delineation, we preferred the amplitude binning method which resulted in a 90% data completeness for a 6 minute acquisition. Amplitude binning can lead to 4DMRI that can be implemented in the clinical workflow using commercially available sequences. The superior contrast compared to the gold standard in radiotherapy, 4DCT, provides the opportunity for better target definition for treatment.We demonstrated the reconstruction of accurate respiratory-correlated 4DMRI as an alternative for 4DCT by creating fast T2-weighted 4D volumetric images.