Self-navigated three-dimensional cardiac T₂ mapping at 3T

Background

Cardiac T₂ mapping using a variable T₂ preparation module (T₂Prep) has recently gained attention for its ability to quantify the extent of edema (Giri, JCMR 2009). Due to time constraints, the T₂ maps are commonly acquired as one or several two-dimensional slices, while the underlying pathology has a three-dimensional (3D) structure. The next logical step would therefore be to exploit recent hardware and software advances to directly acquire 3D T₂ maps. To this end, we tested the feasibility of using a self-navigated 3D radial acquisition with a variable T₂Prep for 3D T₂ mapping at 3T.

Methods

Approval was obtained from the institutional review board. A 3D self-navigated undersampled balanced steady-state free precession (bSSFP) sequence (TR/TE=2.6/1.33ms, matrix 128³, flip angle 70°) with a spiral phyllotaxis radial 3D trajectory (Piccini, MRM 2011) was implemented on a 3T clinical system (Skyra, Siemens AG). This self-navigated pulse sequence allows free breathing acquisitions with 100% scan efficiency, while ECG triggering every 2 heartbeats and TE_T2Prep=60/30/0ms allow for a total acquisition time of ~18min with an isotropic spatial resolution of (1.7mm)³. The datasets were registered using 3D affine registration (Studholme, Med Image Anal 1996). Through Bloch equation simulations, the heart-rate-dependent T₁ -relaxation-related offset in the T₂-fitting equation was ascertained. Subsequently, the validity and accuracy of the T₂ fitting was tested in a phantom whose "true" T₂ values were previously determined. The in vivo robustness of the T₂ determination was then tested in 9 healthy adult subjects. Finally, the sequence was applied for the detection of edema in a 75-year-old male infarct patient after revascularization of his proximal left circumflex.

Results

The Bloch equation simulations of the pulse sequence demonstrated that the input T₂ value could be accurately fitted from the magnetization M with the equation [M=M ₀e ^-TET2Prep/T2 +0.08M₀], while the fitted T₂ had only a ~3% variation over the common range of heart rates (Fig.1A). The phantom T₂ maps demonstrated high homogeneity and fitting accuracy with the 3D sequence matching the ‘true' value to within 1% (Fig.1B). The volunteer study (Fig.2A-C) suggested good agreement with previously reported T₂ values at T₂=39.3±3.9ms (Van Heeswijk, JACC Imaging 2012, in press). A region of significantly elevated T₂ (60.4±9.1 vs. 41.0±4.5ms) was identified in the patient in the infero-lateral myocardium of the left ventricle (Fig.2D,E), consistent with the findings on X-ray coronary angiography.

Figure 1

A) Bloch equation simulation demonstrating that the dependence of the T₂ fit of the magnetization on the heart rate due to varying T₁ relaxation is relatively low (between 43 and 46ms, a variation of 3%), while the "true" input T₂ was 45ms. B) T₂ map of a phantom that approximates arterial blood and myocardium. The T₂ values of the two ‘myocardium' compartments (turquoise) are very similar at 35.3±2.1ms and 35.5±2.4ms and within 1% of the "‘true" T₂ value of 35.6ms.

Figure 2

A-C) Axial, sagittal and coronal multi-planar reformatted T₂ maps through the LV of a healthy volunteer. The myocardium is well defined and T₂=41.3±2.1ms. D) A sagittal T₂ map of a patient with a subacute myocardial infarction demonstrates elevated T₂=62.4±9.2ms in the inferior and infero-lateral segments (arrows). E) 3D segmented LV at a sub-endocardial surface as seen from a posterior position, with a clearly visible inferior and infero-lateral infarction.

Conclusions

The proposed technique provides an easy and time-efficient way to obtain accurate isotropic T₂ maps of the whole heart. Accurate T₂ values were obtained in the phantom, while those in volunteers are consistent with previously reported values. The preliminary patient study demonstrated elevated T₂ in the infarcted region as expected.

Funding

Foundation Emma Muschamp