Characterization of infarct heterogeneity can inform therapeutic strategies for ventricular tachycardia and the management of patients with prior myocardial infarction (MI)^3,4. Two-dimensional (2D) MCLE² images offer better visualization of MI than 2D late gadolinium enhancement (LGE). However, current MR images either with 2D LGE or 2D MCLE provide an inferior spatial resolution of 1.6-2.0mm in-plane with a slice thickness of 5-8mm in the clinical setting. Accurate characterization of the heterogeneous tissue (gray zone) depends on the spatial resolution of the image¹. We recently proposed a novel method COSEP to reconstruct MCLE images with fine details from a highly accelerated dataset and have shown its capability to produce high isotropic resolution in a preclinical study⁵. This work was performed to evaluate the use of COSEP in the characterization of gray zone in patients with MI by using accelerated 3D MCLE with isotropic spatial resolution.

The study was approved by our institutional research ethics board and informed consent was obtained by all subjects. Three patients (average age of 45±4 years, average heart rate of 66±2 bpm) with prior MI were included in this study.

Imaging was performed starting 10 min after injection of 0.2mmol/kg Gadolinium-DTPA on a GE 1.5T scanner with a 30-channel cardiac coil array. 2D LGE and accelerated 3D MCLE were performed on each patient and were prescribed to cover the infarct region in the left ventricle. The parameters for 2D LGE were as follows: FOV=35cm; acquisition matrix=256x160; TR/TE=5.9/1.6ms; flip angle=15^◦; slice thickness=8mm. Each slice was acquired within a single breath-hold. The accelerated 3D MCLE acquisition used the following parameters: acquisition matrix=160x160x10; slab thickness=2.2cm; FOV=35cm; flip angle=45^◦; TR/TE=3.2/1.4ms. The 3D MCLE slab with isotropic spatial resolution of 2.2mm was acquired within a single breath-hold. The undersampled k-space was prospectively acquired using Variable Density Poisson-disk Sampling (VDPDS) patterns at a net acceleration of 5. The VDPDS pattern was different for each inversion time.

In the COSEP reconstruction, the compressed sensing (CS)⁶ framework allows stable reconstruction from a highly undersampled dataset; the temporal signal-relaxation characteristic vectors are extracted using principal component analysis on a training set generated with a pre-determined parametric model, and are then utilized to recursively transform spatiotemporal signal vectors to spatial principal component (PC) maps, facilitating reduction of noise and incoherent artifacts; weighted total variation regularization⁷ is locally applied to preserve anatomical edges in the selected regions on the spatial PC coefficient maps. The mathematical detail for COSEP and the whole reconstruction pipeline were presented in our recent work⁵. The combination of CS and edge preservation could effectively reconstruct fine details in infarcted regions, facilitating accurate characterization of gray zone. For comparison, an alternative CS reconstruction method REPCOM⁸ was also implemented. Data from each coil was processed independently and final MCLE images were generated by using the root-sum-of-squares method. The MCLE image series was then used to obtain a tissue-dependent MR parameter T₁^* map and a steady state signal intensity M_SS map using least squares parametric fitting. With the T₁^* and M_SS maps, a fuzzy c-means clustering algorithm was used for tissue classification⁹.

As seen in FIG. 1, COSEP provides the highest reconstructed spatial resolution on the magnitude image, computed exclusively within the area indicated by the box. Specifically, COSEP preserves well the fine details in the infarct region indicated by the arrow. The reconstructed MCLE image is also shown to have better contrast between infarct and blood than LGE. In FIG. 1D, COSEP presents sharper contrast between infarct and healthy myocardium than REPCOM along the signal intensity profile taken along the dashed line in FIG. 1B.

In FIG. 2, COSEP yields smaller gray zone areas than REPCOM. Specifically, the infarct-myocardium border indicated by the arrow is well delineated by COSEP, while the myocardial pixels inside the little infarct branch are classified as gray zone by REPCOM, likely reflecting partial volume effects.

In FIG. 3, the reduced gray zone area with COSEP vs REPCOM is statistical significant across three patients using a paired Student’s t-test (P = 0.001). Since REPCOM yields relatively blurry anatomical edges in the heterogeneous infarct region, REPCOM likely overestimates the gray zone area, when compared to COSEP.

We have successfully demonstrated that an accelerated MCLE acquisition with COSEP enables high isotropic spatial resolution imaging for improved characterization of infarct heterogeneity in the clinical setting. We have also shown that an isotropic resolution of 2.2mm was achieved within a single breath-hold in a patient study.