Among the spectrum of patients without known disease undergoing elective invasive angiography, nearly 40% do not have coronary artery disease (CAD).¹ It is likely that a significant portion of such patients suffer from ischemic heart disease with underlying coronary microvascular dysfunction (CMD).² Despite intense interest and several recent advancements, reliable diagnosis of CMD on the basis of stress first-pass perfusion (FPP) cardiac MRI is an ongoing challenge. We hypothesized that high-resolution systolic FPP imaging can detect diffuse stress-induced subendocardial defects and transmural perfusion gradients (TPGs) consistent with CMD in a swine model of diet-induced diabetes with no obstructive CAD. To this end, we developed, optimized, and tested a new high-resolution FPP method with hybrid 2D/3D excitation capable of imaging all myocardial slices at the end-systolic phase. To develop a new steady-state "continuous" acquisition scheme for end-systolic first-pass perfusion myocardial MRI at 1.2x1.2 mm² in-plane resolution, and test its effectiveness for reliable detection of subendocardial perfusion defects in a large animal model of coronary microvascular dysfunction.Yucatan mini-pigs (n=8 males) were fed either a high-fat high-sugar diet (n=4 “HFHS” group) or a normal chow diet (n=4 aged-matched “control” group) for ~20 weeks. Compared to controls, the HFHS pigs were obese (68±8 kg vs. 45±7), and had abnormally elevated fasting glucose (177±19 mg/dL vs. 94±12) and insulin levels, indicating early-stage type-2 diabetes and expected to have CMD based on previous validation studies.³ Obstructive CAD was ruled out in all pigs using invasive coronary angiography on the day of the MRI study, consistent with normal serum lipid levels in all animals. There was no difference between baseline arterial systolic blood pressure between the two groups measured during anesthesia (112±8 mmHg vs. 109±6) suggesting absence of hypertension in the HFHS group. Vasodilator stress/rest FPP data (adenosine dose: 210 μg/kg/min) was acquired using a novel T1-weighted steady-state ungated pulse sequence with spoiled GRE readouts (without saturation recovery preparation). Fig. 1 describes the proposed hybrid 2D/3D acquisition scheme in comparison to (a) the conventional magnetization-prepared approach, and (b,c) two recently-introduced “continuous” ungated 3D⁴ and 2D⁵ methods, respectively. The developed pulse sequence used a 16-degree flip angle for both 2D and 3D pulses, which was determined based on phantom experiments to achieve maximal contrast-to-noise ratio (echo spacing = 2.5 ms, acquired in-plane resolution: 1.2x1.2 mm²). Images were reconstructed using a recently-validated compressed sensing (CS) scheme for accelerated radial FPP imaging,⁵ based on a data-adaptive “reference constrained” transform, described in Fig. 2(a). An overview of the reconstruction scheme is provided in Fig. 2(b). The “real-time” navigator (Step 1) is generated by low-resolution reconstruction of the mid ventricular slice at a frame rate of 30 frames/s. Following systolic self-gating (automatic assignment of end-systolic time stamps), a set of sliding window “reference frames” are reconstructed using conjugate-gradient SENSE for each of the 3 slices and used to perform end-systolic time-resolved reconstruction for all 3 slices. The CS scheme incorporated apodization to minimize the subendocardial dark-rim artifact. TPG analysis was performed according to a previously established approach.⁶Representative myocardial perfusion images for a HFHS pig are shown in Fig. 3(a) demonstrating presence of global adenosine-induced subendocardial perfusion defects at the end-systolic phase. The zoomed-in image in Fig. 3(b) shows the highly-resolved subendocardium with >10 pixels covering the transmural distance. Visual assessment of stress FPP images (2 blinded readers in consensus) showed normal perfusion in the control group but a delayed wash-in of contrast in the subendocardium vs. subepicardium for all HFHS pigs. Quantitative TPG analysis, shown in Fig. 4, demonstrated a significantly higher mean TPG across all myocardial segments in HFHS pigs compared to controls (slice-averaged value: 24%±5% vs. 5%±3%), consistent with the qualitative results (example shown in Fig. 3).

The developed steady-state FPP imaging approach with hybrid 2D/3D acquisition is the first method to enable reconstruction of all myocardial slices at the end-systolic phase with an unprecedented 1.2x1.2 mm² resolution without the need for ECG gating. The significantly higher TPG for the HFHS pigs is indicative of impaired myocardial perfusion reserve in the subendocardial layer (corresponding to the location of the microvascular network) during stress, consistent with CMD. In conclusion, the presented results demonstrate that subendocardial ischemia and stress-induced TPGs can be visually detected in the absence of obstructive CAD using the developed high-resolution end-systolic FPP method. The combination of high in-plane resolution, end-systolic imaging of all slices, and dark-rim-minimized reconstruction enables reliable detection of perfusion defects at the subendocardial layer. Our results suggest that this methodology may be a promising approach for accurate diagnosis of CMD in clinical settings.