Synopsis

Keywords: Relaxometry, New Signal Preparation Schemes, T1ρ, Dark-blood, adiabatic

T_1ρ-mapping is emerging as a promising, contrast-free alternative to LGE for assessment of myocardial viability. However, high blood T_1ρ values can obfuscate endocardial scar. In this work, we propose a dark-blood adiabatic T_1ρ preparation to suppress the blood signal and improve depiction of the blood-myocardium interface. A slice-selection gradient is added to an odd number of adiabatic full passage pulses to achieve blood inversion outside the imaging slab. Phantom results show that DBT_1ρ yields unbiased estimation of T_1ρ time. In vivo, thorough blood suppression is achieved for the trade-off against a moderate increase in DBT_1ρ variance.

Introduction

Late gadolinium enhancement (LGE) remains the clinical gold standard for the assessment of myocardial viability. In LGE, blood appears as bright as the myocardium, thus the detection of subendocardial lesions can be challenging. To alleviate this, dark-blood (DB) LGE has been recently proposed, nulling the healthy myocardium and attenuating the blood signal[1]. T_1ρ mapping is an emerging non-contrast alternative to LGE for the discrimination of healthy and infarcted myocardium, but faces similar problems at the sub-endocardium [2-5]. The blood pool is characterized by high T_1ρ, potentially obfuscating endocardial scar. In this study, we propose a novel adiabatic T_1ρ preparation to achieve dark-blood contrast in T_1ρ-adiab mapping for improved depiction of the blood-myocardial interface.

Methods

Previously proposed bright-blood adiabatic spin-lock (SL) preparations were used to acquire reference T_1ρ-adiab maps (RefT_1ρ-adiab)[6]. These consisted of trains of even-numbered hyperbolic secant (HS) adiabatic full passage (AFP) pulses (τ_HS=30ms, β=6.9, f_max=450Hz)(Fig 1a). DB contrast was achieved by playing a 880μT/m slice-selection gradient for all HS pulses in the preparation train, except for the last. This way, an even number of AFPs is applied to the labeling slab, inducing T_1ρ-adiab decay in the imaging slice. Outside the slab, only the last AFP is experienced, leading to magnetization inversion (Fig 1b). A delay (δ) was added between the preparation and the acquisition to allow for the inverted blood to flow into the imaging slab. The preparation duration was maximized for each volunteer, depending on the subject’s heart rate (δ=320-750ms). Bloch simulations and phantom imaging with δ=0-500ms were performed to assess the influence of the delay on the T_1ρ-adiab variance. T_1ρ-adiab mapping was performed on a 3T scanner (Philips Ingenia) with a 14s breath-hold cardiac-triggered snap-shot bSSFP sequence (Fig 1c). Three T_1ρ-prepared images (𝜏_SL=60, 120, 180ms for 2, 4, HS pulses) were acquired, interleaved by 4s Mz recovery pauses and followed by a saturation-prepared image[6]. Other sequence parameters were: resolution=2x2x8mm3, FOV=220x220mm2, TE/TR=0.76/1.90ms. All T_1ρ-adiab maps were generated using a three-parameter model[7]. Precision (CV_p) and repeatability (CV_r) were assessed in the T1MES phantom[8] for RefT_1ρ-adiab and DBT_1ρ-adiab with varying δ. Ten repetitions were acquired for each case. Finally, the sequence was tested in four healthy volunteers (1 female, 3 males, 27±5y/o). For each volunteer, 3 repetitions of mid short-axis (SAX) were acquired to assess repeatability, in addition to single apical SAX, basal SAX and 4 chambers views. CV_p, CV_r and inter-subject variability (CVs), were investigated in manually segmented myocardium ROIs.

Results

T_1ρ-adiab maps obtained with RefT_1ρ-adiab and DBT_1ρ-adiab show good agreement for all preparation delays (Fig 2, myocardium-like vial: RefT_1ρ-adiab: 184.73±7.83ms, DBT_1ρ-adiab with δ=500ms: 188.68±18.73ms). Precision and repeatability were comparable to RefT_1ρ-adiab when no preparation delay was used (Fig 3b-c, DBT_1ρ-adiab with δ=0ms: CV_p=4.24±0.49%, CV_r=3.72% vs RefT_1ρ-adiab: CV_p=4.00±0.56%, CV_r=2.03%). However, larger variability is observed for increasing δ, due to magnetization relaxation after the preparation (Fig 3a, linear regression: slope=19.61, intercept=7.38, R²=0.81, p-value<<0.05). In-vivo testing of the preparation with slice-selection gradient along the phase and frequency-encoding dimensions, shows sharp delineation of an inversion and a pass band (Fig 4a). When being played along the slice-selection direction, thorough blood suppression is observed (Fig 4b). For subjects with high heart rates, where only short δ was possible, complete suppression of the blood is compromised (Fig 4b). However, the blood near the myocardial interface remains thoroughly nulled. In vivo, homogenous and artifact-free T_1ρ-adiab quantification is observed with both RefT_1ρ-adiab and DBT_1ρ-adiab (Fig. 5). No significant difference was found between myocardial DBT_1ρ-adiab values and the reference (DBT_1ρ-adiab=197.15±42.31ms, RefT_1ρ-adiab=202.03±26.45ms, p>0.05). DBT_1ρ-adiab maps show compromised precision, repeatability and inter-subject variability (DBT_1ρ-adiab CV_p=24.62±5.10%, CV_r=12.23±3.84%, CV_s=14.67%, vs RefT_1ρ-adiab CV_p=12.66±2.09%, CV_r=7.16±1.90%, CV_s=11.28%, p<0.05).

Discussion

In this study we explored the use of dark-blood contrast in adiabatic T_1ρ mapping in the myocardium. Our results show that DBT_1ρ-adiab maps yield quantification and map quality comparable to conventional RefT_1ρ-adiab, while achieving thorough blood signal suppression. Increasing preparation delays resulted in improved blood suppression by allowing nulling of the inverted blood magnetization and its inflow to the pass band, albeit at the trade-off against reduced precision. In the presence of short preparation delays, increased imaging flip angles could be used to drive the blood signal closer to the zero crossing, for the trade-off against decreased contrast sensitivity. The proposed implementation includes the acquisition of a saturation-prepared image with equal delay. This allows to capture both the magnetization relaxation, as well as the effect of the imaging readout. In combination with a 3 parameter fit this enables unbiased DBT_1ρ-adiab estimation, irrespective of the preparation delay.

Conclusions

DBT_1ρ-adiab preparations enable thorough blood signal suppression in adiabatic T_1ρ mapping, at a moderate increase in T_1ρ-adiab quantification variance. Thus, DBT_1ρ-adiab forms a promising alternative to conventional T_1ρ mapping with the potential to enhance visualization of subendocardial scar.

Acknowledgements

Funding:4TU Precision Medicine program, NWO Start-up STU.019.024, ZonMW OffRoad grant 04510011910073.