Abdominal simultaneous 3D water T1 and T2 mapping using a free-breathing Cartesian acquisition with spiral profile ordering

Abstract

Purpose: To develop a method for abdominal simultaneous 3D water $T_1$ ( $w{T}_1$ ) and $T_2$ ( $w{T}_2$ ) mapping with isotropic resolution using a free-breathing Cartesian acquisition with spiral profile ordering (CASPR) at 3 T.

Methods: The proposed data acquisition combines a Look-Locker scheme with the modified BIR-4 adiabatic preparation pulse for simultaneous $w{T}_1$ and $w{T}_2$ mapping. CASPR is employed for efficient and flexible k-space sampling at isotropic resolution during free breathing. The imaging pipeline includes subspace reconstruction, water-fat separation, and $B_0$ -specific dictionary matching. The proposed method was validated in a water-fat relaxometry phantom using spin echo-based reference techniques and was compared with MOLLI $T_1$ and GRASE $T_2$ mapping in 10 volunteers. The method's flexibility was assessed at isotropic resolutions of 2.5, 3, and 3.5 mm, with corresponding scan times of 7:19, 5:23, and 3:52 min. Additionally, the method was applied to 9 oncological patients with abdominal pathologies.

Results: Phantom experiments demonstrated good agreement between the proposed method and spin echo-based reference techniques across a wide range of $w{T}_1$ and $w{T}_2$ values. The volunteer study assessed in vivo quantification performance and demonstrated the flexibility of the proposed method at different spatial resolutions. The 3 mm method was successively performed in patients with abdominal pathologies, whereby lesions with diameters below 1 cm could be assessed.

Conclusion: The proposed Look-Locker-based method using CASPR enables efficient, flexible, and simultaneous $w{T}_1$ and $w{T}_2$ mapping at isotropic resolution in a fixed scan time during free breathing.

Keywords: Look‐Locker; isotropic resolution; liver; relaxometry; water‐fat separation.

FIGURE 1

Schematic overview of (A) the proposed acquisition scheme and (B) the image reconstruction and parameter estimation pipeline. (A) The acquisition consists of four blocks, each containing an adiabatic preparation pulse followed by a Look‐Locker readout. A golden‐step CASPR trajectory was applied varying every shot the sampled profiles. (B) The undersampled raw k‐space data were reconstructed using a soft‐gated, total‐variation (TV)‐regularized subspace reconstruction. The subspace reconstruction was followed by water‐fat separation and $\mathrm{w}{T}_1$ and $\mathrm{w}{T}_2$ estimation using a $B_0$‐specific dictionary approach.

FIGURE 2

Validation of the proposed method at 3 mm isotropic resolution in a water‐fat relaxometry phantom. The $\mathrm{w}{T}_1$ and $\mathrm{w}{T}_2$ estimates of the proposed method show excellent agreement with Dixon inversion‐recovery spin‐echo (Dixon IR‐SE) $\mathrm{w}{T}_1$ and Dixon spin‐echo (Dixon SE) $\mathrm{w}{T}_2$ mapping. In addition, the performance of the proposed method is compared to vendor‐implemented MOLLI $T_1$ and GRASE $T_2$ mapping techniques.

FIGURE 3

Visual comparison of $\mathrm{w}{T}_1$ and $\mathrm{w}{T}_2$ maps from the proposed 3D method (3 mm isotropic resolution) with MOLLI $T_1$ and GRASE $T_2$ in three volunteers. The proposed method is presented in both axial and coronal views, while the 2D MOLLI and GRASE maps are shown in the axial plane.

FIGURE 4

Quantitative comparison of the proposed method with MOLLI $T_1$ and GRASE $T_2$ mapping for the volunteer study. ROIs were placed in the liver and back muscle for all ten volunteers. Above is the linear correlation for $T_1$ and $T_2$. Below are Bland‐Altman plots comparing the differences between the methods. The lower $T_1$ values for MOLLI and the higher $T_2$ values for GRASE agree with the results in the phantom. The phantom references, Dixon IR‐SE, and Dixon SE, cannot be acquired in vivo due to the long acquisition times.

FIGURE 5

Visual comparison of the proposed method at 2.5, 3, and 3.5 mm isotropic resolution for the estimated $\mathrm{w}{T}_1$ and $\mathrm{w}{T}_2$ maps, as well as PD‐like water images. Axial (left) and coronal (right) views are shown for a representative volunteer.

FIGURE 6

Reproducibility analysis of the proposed method at different spatial resolutions without repositioning in five representative volunteers. ROIs in the liver, pancreas, spleen, and muscle are compared at (a) 2.5 mm and (b) 3.5 mm isotropic resolution against the 3 mm acquisition. The results show a strong correlation across all analyses, with the largest variations observed in pancreatic ROIs.

FIGURE 7

$B_0$ inhomogeneities can lead to quantification errors if they are not taken into account during dictionary matching. Proposed $\mathrm{w}{T}_1$ and $\mathrm{w}{T}_2$ maps are compared with and without considering $B_0$ in dictionary matching. The estimated field map from the data is shown as a reference. Strong $B_0$ inhomogeneities are observed in a volunteer in the upper liver region near the lung (top) and in a patient with a metal clip, leading to signal extinction (bottom). The regions are indicated by arrows highlighting primarily underestimated $\mathrm{w}{T}_1$ without $B_0$‐specific dictionary matching.

FIGURE 8

Proposed $\mathrm{w}{T}_1$ and $\mathrm{w}{T}_2$ mapping in a patient with ischemic cholangiopathy. $3$ mm $\mathrm{w}{T}_1$ and $\mathrm{w}{T}_2$ maps, as well as PD‐like water images, are presented in axial and coronal views and compared to $T_2$‐weighted and native Dixon $T_1$‐weighted images from the clinical protocol. Red arrows indicate a liver region with lower $\mathrm{w}{T}_2$, co‐localized with signal reduction on the clinical $T_2$‐weighted images.

FIGURE 9

Proposed $\mathrm{w}{T}_1$ and $\mathrm{w}{T}_2$ mapping in a patient with hepatocellular carcinoma (HCC) following selective internal radiotherapy (SIRT), demonstrating diffuse post‐radiogenic parenchymal changes. $3$ mm $\mathrm{w}{T}_1$ and $\mathrm{w}{T}_2$ maps, as well as PD‐like water images, are shown in axial and coronal views. As a reference from the clinical protocol, $T_2$‐weighted, native Dixon $T_1$‐weighted, and Dixon $T_1$‐weighted images from the portal venous (PV) phase are included. Arrows indicate two small lesions ($<1$ cm in diameter) visible in the clinical $T_2$‐weighted and contrast‐enhanced $T_1$‐weighted images. The lesions are well depicted in the proposed $\mathrm{w}{T}_1$ and $\mathrm{w}{T}_2$ maps. The clinical native Dixon $T_1$‐weighted scan required two repetitions due to motion artifacts during the breath‐hold. The slice position is marked by a red line in the coronal view of the water image.