Quantifying spatial and dynamic lung abnormalities with 3D PREFUL FLORET UTE imaging: A feasibility study

Abstract

Purpose: Pulmonary MRI faces challenges due to low proton density, rapid transverse magnetization decay, and cardiac and respiratory motion. The fermat-looped orthogonally encoded trajectories (FLORET) sequence addresses these issues with high sampling efficiency, strong signal, and motion robustness, but has not yet been applied to phase-resolved functional lung (PREFUL) MRI-a contrast-free method for assessing pulmonary ventilation during free breathing. This study aims to develop a reconstruction pipeline for FLORET UTE, enhancing spatial resolution for three-dimensional (3D) PREFUL ventilation analysis.

Methods: The FLORET sequence was used to continuously acquire data over 7 ± 2 min in 36 participants, including healthy subjects (N = 7) and patients with various pulmonary conditions (N = 29). Data were reconstructed into respiratory images using motion-compensated low-rank reconstruction, and a 3D PREFUL algorithm was adapted to quantify static and dynamic ventilation surrogates. Image sharpness and signal-to-noise ratio were evaluated across different motion states. PREFUL ventilation metrics were compared with static ¹²⁹Xe ventilation MRI.

Results: Optimal image sharpness and accurate ventilation dynamics were achieved using 24 respiratory bins, leading to their use in the study. A strong correlation was found between 3D PREFUL FLORET UTE ventilation defect percentages (VDPs) and ¹²⁹Xe VDPs (r ≥ 0.61, p < 0.0001), although PREFUL FLORET static VDPs were significantly higher (mean bias = -10.1%, p < 0.0001). In diseased patients, dynamic ventilation parameters showed greater heterogeneity and better alignment with ¹²⁹Xe VDPs.

Conclusion: The proposed reconstruction pipeline for FLORET UTE MRI offers improved spatial resolution and strong correlation with ¹²⁹Xe MRI, enabling dynamic ventilation quantification that may reveal airflow abnormalities in lung disease.

Keywords: FLORET; PREFUL; UTE; hyperpolarized 129Xe MRI; lung MRI; motion compensation; ventilation imaging.

FIGURE 1

Respiratory binning workflow. (A) Waveform collected from respiratory bellows measurement (only first 20 000 excitations; ˜1 min shown). (B) Hilbert‐transform of respiratory bellows waveform to separate each respiratory cycle into phase position −$\pi$ to $\pi$. Transformed data were binned into 24 windows containing 10 000 excitations each. Different colors refer to different respiratory bins. (C) The degree of overlap within each bin. Small white lines in (C) correspond to excluded excitations that fell beyond ±2 standard deviations of the mean respiratory bellows measurement.

FIGURE 2

Three‐dimensional FLORET (fermat‐looped orthogonally encoded trajectories) reconstruction pipeline for this study. In chronological order: (1) Excitation indices are binned according to position in respiratory cycle and applied to raw k‐space data and trajectories; (2) sensitivity maps and k‐space preconditioners are calculated; (3) conjugate‐gradient sensitivity encoding (CG‐SENSE) is explored to initialize $N$ images corresponding to each respiratory phase; (4) motion field from initial images is estimated, iterated through the motion‐compensated low‐rank reconstruction (MoCoLoR) algorithm using the initial motion fields as inputs; and (5) reconstruction is complete.

FIGURE 3

Comparison of different regularization parameters and their influence on image quality of ventilation images derived from three‐dimensional PREFUL (phase‐resolved functional lung) FLORET (fermat‐looped orthogonally encoded trajectories) ultrashort echo time. Overview of morphological end‐expiration (left) and regional ventilation (RVent; right) maps derived using different reconstruction settings (regularization parameter, λ_L) for a healthy participant (male, 11 years old). Note the superior Sobel image sharpness (∇) of ventilation images and improved qualitative sharpness around diaphragm as well as well‐defined vessels of morphological images for image reconstructions with λ_L = 0.005 and 0.01. VDP, ventilation defect percentage.

FIGURE 4

Comparison of static three‐dimensional PREFUL (phase‐resolved functional lung) FLORET (fermat‐looped orthogonally encoded trajectories) ultrashort echo time MRI (first–second row) and breath‐hold ¹²⁹Xe ventilation MRI (third row). Four study participants are depicted: healthy volunteer (first column, female, 13 years old), subject with bronchiolitis obliterans syndrome (BOS; second column, male, 24 years old), subject with cystic fibrosis (CF; third column, female, 25 years old), and subject with lymphangioleiomyomatosis disease (LAM; fourth column, female, 39 years old). For both measurements, the whole‐lung ventilation defect percentage (VDP) values are listed below the respective ventilation map.

FIGURE 5

Comparison of three‐dimensional PREFUL (phase‐resolved functional lung) FLORET (fermat‐looped orthogonally encoded trajectories) ultrashort echo time MRI–derived ventilation cycle. Healthy volunteer (first row, male, 13 years old), cystic fibrosis (CF; second row, female, 25 years old) patient, bronchiolitis obliterans syndrome (BOS; third row, male, 24 years old) patient, and lymphangioleiomyomatosis (LAM; fourth row, female, 39 years old) patient are presented. Note homogenous ventilation cycle for the healthy volunteer and no ventilation filling for consolidation in the right lung of the CF patient (cyan arrows). For both BOS and LAM subjects, the ventilation dynamics are much more heterogeneous. Dynamic filling asynchronization is indicated with yellow arrows for BOS and green arrows for LAM subjects, suggesting abnormal airflow related to airway narrowing.

FIGURE 6

Comparison of static and dynamic three‐dimensional (3D) PREFUL (phase‐resolved functional lung) FLORET (fermat‐looped orthogonally encoded trajectories) ultrashort echo time (UTE) MRI (first–fourth row) and breath‐hold ¹²⁹Xe ventilation MRI (fifth row). Healthy volunteer (first column; male, 25 years old) subject with bronchiolitis obliterans syndrome (BOS; second column, male, 24 years old), subject with cystic fibrosis (CF; third column, male, 25 years old), and subject with lymphangioleiomyomatosis disease (LAM; fourth column, female, 39 years old) are depicted. The overlapping hypoventilation areas depicted in static regional ventilation of 3D PREFUL FLORET UTE (second row) and ¹²⁹Xe ventilation imaging (fifth row) are marked with cyan arrows. In contrast, overlapping ventilation abnormalities derived with dynamic 3D PREFUL FLORET UTE parameters (third–fourth row) and ¹²⁹Xe breath‐hold imaging are marked with red arrows. CM, cross‐correlation metric; FVL, flow‐volume loop; RVent, regional ventilation; VDP, ventilation defect percentage; VTTP, ventilation time‐to‐peak.

FIGURE 7

Correspondence of ventilation defect percentage (VDP) values derived by three‐dimensional (3D) PREFUL (phase‐resolved functional lung) FLORET (fermat‐looped orthogonally encoded trajectories) ultrashort echo time MRI and ¹²⁹Xe imaging. Bland–Altman (A,C) and correlation analysis (B,D) of global VDP values derived using 3D PREFUL MRI postprocessing and ¹²⁹Xe MRI. Study groups are visually marked with distinct colors (green, healthy volunteers; pink, patients with cystic fibrosis [CF]; blue, patients with lymphangioleiomyomatosis [LAM]; orange, patients with bronchiolitis obliterans syndrome [BOS]; purple, patients post–hematopoietic stem cell transplantation [Post‐HSCT]; cyan, patient with neuroendocrine cell hyperplasia of infancy [NEHI]). cross‐correlation metric; FVL, flow‐volume loop.