Breath-held high-resolution cardiac T₂ mapping with SKRATCH

Background

Several cardiac T₂ mapping techniques with varying T₂ preparation (T₂Prep) times have been proposed for the quantification of cardiac edema [1–3]. Among these, radial T₂ mapping, which is robust to motion artifacts, suffers from a low signal-to-noise ratio (SNR) caused by the undersampling of the k-space periphery and by its density compensation function (DCF) (Fig. 1a). However, since the contrast of an image is mainly determined by the center of its k-space, the T₂-weighted images can share their k-space periphery using the KWIC (K-space Weighted Image Contrast) filter (Fig. 1b) to reduce undersampling artifacts [4]. This allows for higher undersampling (Fig. 1c) and thus for a decrease in acquisition time [5].

Figure 1

Schematic overview of the KWIC filter. a. A radial k-space sampling pattern shown below its DCF along one radial line. The DCF is used to weigh the k-space points. b. Three similar k-spaces that share their periphery through the KWIC filter, thus increasing the local sampling density and decreasing the local weight attributed by the DCF. The radii outside of which data were added were defined through the Nyquist criterion. c. An undersampled KWIC-filtered k-space. While the number of lines has decreased, the periphery of k-space still has a higher sampling density than the standard radial k-space in a.

We demonstrated that navigator-gated KWIC-filtered cardiac T₂ mapping (Shared K-space RAdial T₂ Characterization of the Heart, SKRATCH) enables a considerable decrease in acquisition time while maintaining the T₂ precision [5]. The goal of this study was to extend this approach to a short breath-held high-resolution T₂ map acquisition and to compare its performance to navigator-gated T₂ mapping.

Methods

The novel breath-held SKRATCH protocol consisted of a GRE sequence with a continuously increasing golden-angle radial acquisition. This ensured a unique k-space trajectory for all 64 lines of each of the 4 T₂Prep durations (0/30/45/60 ms), pixel size of 1.2 × 1.2 × 8 mm³ and a total duration of 7 heartbeats. As reference, a navigator-gated radial cardiac T₂ mapping GRE sequence was acquired with 3 T₂Prep durations (0/30/60 ms), 308 lines/image and a pixel size of 1.25 × 1.25 × 5 mm³ [3]. Images were acquired at 3T (Magnetom Prisma, Siemens Healthcare) in 17 healthy volunteers at the same midventricular short-axis orientation with both protocols. The T₂ maps were segmented according to the AHA guidelines [6]. The mean T₂ value (μ_T2) and the relative standard deviation (σ_R = standard deviation/ μ_T2) of each segment as well as the myocardial area were calculated and tested for significant differences. The SKRATCH T₂ map was acquired twice in 11 of the volunteers for Bland-Altman reproducibility analysis.

Results

The SKRATCH T₂ maps had average values of 39.9 ± 4.4 ms, while those of the reference T₂ maps were 39.1 ± 3.1 ms (p = 0.04, Fig. 2a-c). σ_R increased from 8 ± 2% for the standard T₂ maps to 11 ± 2% for the SKRATCH T₂ maps (p < 0.001). The myocardial area decreased from 643 ± 155 to 585 ± 121 pixels for the SKRATCH T₂ maps (a 10% decrease, p = 0.008). The repeatability analysis resulted in a confidence interval of ± 3.09 ms (Fig. 2d).

Figure 2

A comparison of navigator-gated and breath-held high-resolution T ₂ maps in healthy volunteers. a,b. The standard navigator-gated T₂ map and breath-held SKRATCH T₂ map respectively. Note that the maps are homogeneous and have similar myocardial surface available for analysis. The color bar indicates the T₂ relaxation time in ms. c. The mean T₂ values and standard deviations of the 17 healthy volunteers show a slight increase in standard deviation for the breath-held SKRATCH acquisition. d. The Bland-Altman analysis of the difference in mean T₂ values for 11 volunteers. The dotted line represents the mean with a bias of 0.28, while the continuous lines represent the 95% confidence interval (1.96 × standard deviation).

Conclusions

The SKRATCH T₂ maps were highly similar to the reference high-resolution T₂ maps, while the shortening to breath-hold duration came at the cost of an acceptably small increase in standard deviation and decrease in myocardial area. These encouraging results will need to be validated in future high-resolution studies in patients.