Removal of hyperpolarized 129 Xe gas-phase contamination in spectroscopic imaging of the lungs

Abstract

Purpose: A novel technique is presented for retrospective estimation and removal of gas-phase hyperpolarized Xenon-129 (HP ¹²⁹ Xe) from images of HP ¹²⁹ Xe dissolved in the barrier (comprised of parenchymal lung tissue and blood plasma) and red blood cell (RBC) phases. The primary aim is mitigating RF pulse performance limitations on measures of gas exchange (e.g., barrier-gas and RBC-gas ratios). Correction for gas contamination would simplify technical dissemination of HP ¹²⁹ Xe applications across sites with varying hardware performance, scanner vendors, and models.

Methods: Digital lung phantom and human subject experiments (N = 8 healthy; N = 1 with idiopathic pulmonary fibrosis) were acquired with 3D radial trajectory and 1-point Dixon spectroscopic imaging to assess the correction method for mitigating barrier and RBC imaging artifacts. Dependence of performance on TE, image SNR, and gas contamination level were characterized. Inter- and intra-subject variation in the dissolved-phase ratios were quantified and compared to human subject experiments before and after correction.

Results: Gas contamination resulted in image artifacts similar to those in disease that were mitigated after correction in both simulated and human subject data; for simulation experiments performance varied with TE, but was independent of image SNR and the amount of gas contamination. Artifacts and variation of barrier and RBC components were reduced after correction in both simulation and healthy human lungs (barrier, P = 0.01; RBC, P = 0.045).

Conclusion: The proposed technique significantly reduced regional variations in barrier and RBC ratios, separated using a 1-point Dixon approach, with improved accuracy of dissolved-phase HP ¹²⁹ Xe images confirmed in simulation experiments.

Keywords: artifact correction; hyperpolarized MRI; idiopathic pulmonary fibrosis; lung; spectroscopic imaging; xenon MRI.

Figure 1

Pulse sequence diagram (A) of the 3D interleaved gas and dissolved-phase image acquisition shows the analog-to-digital converter (ADC) readout of two echoes following each of a spectrally selective dissolved-phase and then gas-phase RF excitation. The magnitude of the total gradient along the direction of a 3D radial projection is depicted, including readout and spoiler gradients at the end of each half of the TR. Flyback gradients are positioned after the readout at echo time, TE₁ = 0.9 ms, in preparation for the second echo at TE₂ = 4.2 ms. Gradients are identical in the dissolved-phase and gas-phase portions of each TR, and arrows indicate the echo and component acquired during each readout. A flow chart (B) depicts the steps taken by the proposed correction algorithm. The corresponding equations to each processing step are listed.

Figure 2

Views in a single central slice of raw and corrected images from (A–F) simulated data and (G–L) data acquired in a healthy normal human subject (63 year old male, average RBC/Barrier ratio of 0.3). The lung boundary is drawn in white. Note the homogenously elevated barrier signal (A,G) relative to corrected data (B,H), as well heterogeneous artifacts in the RBC (C,I) largely resolved following correction (D,J). Dissolved-phase data from TE₂ were reconstructed on the gas resonance (Eq. 6) to depict residual gas contamination images before (E,K) and after (F,L) correction, respectively.

Figure 3

Quantitative simulation results depicting the expected theoretical performance of the proposed correction. The residual gas contamination error (as a % of dissolved-phase signal at TE₁), is approximately proportional to dissolved-phase signal remaining at TE₂ (A). However, the error is influenced by the relative phase between the RBC and tissue components of the dissolved-phase signal, evidenced by the local minimum in the residual contamination error near 3.7ms in (A). The image SNR also has no effect on the residual contamination error (B). The confidence intervals of the measures are smaller than the width of the line for SNR > 2. Similarly, residual error appears independent of both the (C) magnitude and (D) phase of the gas contamination itself.

Figure 4

Distributions of the whole lung mean (A,C) and standard deviations (B,D) of barrier and RBC values in the simulated (top row) and human subject (bottom row) data with gas contamination and after correction. The large dispersion in the global mean values caused by the contaminant gas signal is removed in the simulation experiments, and for the Barrier/Gas ratio in the human subject experiments. The regional standard deviation after correction reflects reduced regional heterogeneity in both simulation and human subject results (* indicate P<0.05).

Figure 5

Whole lung histograms and parametric maps of barrier-to-gas ratio from 3 healthy normal human subjects (1 per row). Histograms show whole lung parameter distributions that are narrower and more similar across subjects following the gas contamination correction. Uncorrected parametric maps show numerous regions of defect and hyper-intensity that largely resolve post-correction.

Figure 6

Whole lung histograms and parametric maps of barrier-to-gas ratio from 3 healthy normal human subjects (1 per row). Histograms show whole lung parameter distributions that are narrower and more similar across subjects following the gas contamination correction. Uncorrected parametric maps show numerous regions of defect and hyper-intensity that largely resolve post-correction.

Figure 7

Whole lung histograms (A, B) and parametric maps (C, D) of barrier-to-gas ratio (A, C) and RBC-to-gas ratio (B, D), respectively, from a subject diagnosed with idiopathic pulmonary fibrosis. Regions of defect, which are expected to be present in this disease population, remain but in some cases enhanced, following correction with average values increased overall.