Is synthesizing MRI contrast useful for inter-modality analysis?

Abstract

Availability of multi-modal magnetic resonance imaging (MRI) databases opens up the opportunity to synthesize different MRI contrasts without actually acquiring the images. In theory such synthetic images have the potential to reduce the amount of acquisitions to perform certain analyses. However, to what extent they can substitute real acquisitions in the respective analyses is an open question. In this study, we used a synthesis method based on patch matching to test whether synthetic images can be useful in segmentation and inter-modality cross-subject registration of brain MRI. Thirty-nine T1 scans with 36 manually labeled structures of interest were used in the registration and segmentation of eight proton density (PD) scans, for which ground truth T1 data were also available. The results show that synthesized T1 contrast can considerably enhance the quality of non-linear registration compared with using the original PD data, and it is only marginally worse than using the original T1 scans. In segmentation, the relative improvement with respect to using the PD is smaller, but still statistically significant.

Fig. 1

Sagittal slice of a sample MRI scan from the test dataset: (a) PD-weighted volue, (b) T1-weighted volume, (c) T1 volume synthesized from the PD data using [8], (d) synthesized with [9], (e) synthesized with the exemplar-based method.

Fig. 2

Boxplot of Dice scores in the registration experiment with the coarse (top) and symmetric diffeomorphic (bottom) deformation models. See Section 3.2 for the abbreviations. Color code: magenta = PD registered with MI, red = synthetic T1 from [8], green = synthetic T1 from [9], blue = synthetic T1 from exemplar-based approach, black = ground truth T1. Horizontal box lines indicate the three quartile values. Whiskers extend to the most extreme values within 1.5 times the interquartile range from the ends of the box. Samples beyond those points are marked with crosses. All the differences are statistically significant at p=0.001 (sample size 39 × 8 = 312).

Fig. 3

Boxplot of the Dice scores in the segmentation experiment. See caption of Figure 2 for the abbreviations and color code.