Respiratory self-gated 3DUTE for lung imaging in small animal MRI
2016; Wiley; Volume: 78; Issue: 2 Linguagem: Inglês
10.1002/mrm.26463
ISSN1522-2594
AutoresMarta Tibiletti, Andrea Bianchi, Åsmund Kjørstad, Stefan Wundrak, Detlef Stiller, Volker Rasche,
Tópico(s)Electron Spin Resonance Studies
ResumoMagnetic Resonance in MedicineVolume 78, Issue 2 p. 739-745 NoteFree Access Respiratory self-gated 3D UTE for lung imaging in small animal MRI Marta Tibiletti, Corresponding Author Marta Tibiletti [email protected] Core Facility Small Animal MRI, Ulm University, Ulm, GermanyCorresponding author: Marta Tibiletti, Core Facility Small Animal MRI, Ulm University, Albert-Einstein-Allee 23, 89081, Ulm, Germany. E-mail: [email protected].Search for more papers by this authorAndrea Bianchi, Andrea Bianchi Boehringer Ingelheim Pharma GmbH & Co. KG, Target Discovery Research, In-Vivo Imaging Laboratory, Biberach an der Riss, GermanySearch for more papers by this authorÅsmund Kjørstad, Åsmund Kjørstad Department of Neuroradiology, University Hospital Hamburg-Eppendorf, Hamburg, GermanySearch for more papers by this authorStefan Wundrak, Stefan Wundrak Department of Internal Medicine II, Ulm University, Ulm, Germany.Search for more papers by this authorDetlef Stiller, Detlef Stiller Boehringer Ingelheim Pharma GmbH & Co. KG, Target Discovery Research, In-Vivo Imaging Laboratory, Biberach an der Riss, GermanySearch for more papers by this authorVolker Rasche, Volker Rasche Core Facility Small Animal MRI, Ulm University, Ulm, Germany Department of Internal Medicine II, Ulm University, Ulm, Germany.Search for more papers by this author Marta Tibiletti, Corresponding Author Marta Tibiletti [email protected] Core Facility Small Animal MRI, Ulm University, Ulm, GermanyCorresponding author: Marta Tibiletti, Core Facility Small Animal MRI, Ulm University, Albert-Einstein-Allee 23, 89081, Ulm, Germany. E-mail: [email protected].Search for more papers by this authorAndrea Bianchi, Andrea Bianchi Boehringer Ingelheim Pharma GmbH & Co. KG, Target Discovery Research, In-Vivo Imaging Laboratory, Biberach an der Riss, GermanySearch for more papers by this authorÅsmund Kjørstad, Åsmund Kjørstad Department of Neuroradiology, University Hospital Hamburg-Eppendorf, Hamburg, GermanySearch for more papers by this authorStefan Wundrak, Stefan Wundrak Department of Internal Medicine II, Ulm University, Ulm, Germany.Search for more papers by this authorDetlef Stiller, Detlef Stiller Boehringer Ingelheim Pharma GmbH & Co. KG, Target Discovery Research, In-Vivo Imaging Laboratory, Biberach an der Riss, GermanySearch for more papers by this authorVolker Rasche, Volker Rasche Core Facility Small Animal MRI, Ulm University, Ulm, Germany Department of Internal Medicine II, Ulm University, Ulm, Germany.Search for more papers by this author First published: 23 September 2016 https://doi.org/10.1002/mrm.26463Citations: 13 A.B, and D.S. are employees of Boehringer Ingelheim Pharma GmbH & Co. KG. The authors do no report any other conflict of interest. AboutSectionsPDF ToolsRequest permissionExport citationAdd to favoritesTrack citation ShareShare Give accessShare full text accessShare full-text accessPlease review our Terms and Conditions of Use and check box below to share full-text version of article.I have read and accept the Wiley Online Library Terms and Conditions of UseShareable LinkUse the link below to share a full-text version of this article with your friends and colleagues. Learn more.Copy URL Abstract Purpose To investigate retrospective respiratory gating of three-dimensional ultrashort echo time (3D UTE) lung acquisition in free-breathing rats using k-space center self gating signal (DC-SG) and 3D image-based SG (3D-Img-SG). Methods Seven rats were investigated with a quasi-random 3D UTE protocol. Low-resolution time-resolved sliding-window images were reconstructed with a 3D golden-angle radial sparse parallel (GRASP) reconstruction to extract a 3D-Img-SG signal, whereas DC-SG was extracted from the center of k-space. Both signals were sorted into 10 respiratory bins. Signal-to-noise ratio (SNR) and normalized signal intensity (NSI) in lung parenchyma, image sharpness, and lung volume changes were studied in the resulting images to show feasibility of the method. An algorithm for bulk movement identification and removal was implemented. Results Three-dimensional Img-SG allows reconstruction of different respiratory stages in all acquired datasets, showing clear differences in diaphragm position and significantly different lung volumes, SNR, and NSI in lung parenchyma. Improved sharpness in expiration images was observed compared to ungated images. DC-SG did not result in clear different diaphragm position in all cases. Bulk motion removal improved final image sharpness. Conclusion Low-resolution 3D GRASP reconstruction allowed for extraction of an effective gating signal for 3D-Img-SG. The DC-SG method did not work in cases for which respiratory frequencies were inconsistent. Magn Reson Med 78:739–745, 2017. © 2016 International Society for Magnetic Resonance in Medicine INTRODUCTION Imaging of lung parenchyma using proton MRI is particularly challenging 1. The low proton density in lung tissue causes intrinsic low MR signal compared to other organs 2, and the multiple water–air interfaces in lung parenchyma cause rapid signal loss, resulting in a very short T2* well below 500 µs at high field strength. Furthermore, respiratory and cardiac motion must be properly considered, and thus further complicate the application of MRI for lung parenchyma imaging. Ultrashort echo time (UTE) MR techniques have been successfully applied for imaging of lung parenchyma. As a radial center-out method, UTE can be used for two-dimensional (2D) (echo time (TE) about 200 to 300 µs 3, 4) and 3D (TE as short as 8 µs, 5, 6) imaging, with excellent performance for direct lung parenchyma visualization 7. Although UTE, as a radial imaging technique, is intrinsically robust to motion artifacts 8, respiratory compensation is still desirable to maximize image quality in high-resolution imaging 9. Synchronization of the image acquisition to the breathing cycle can be achieved with external sensors 10. Alternative methods aim to extract motion information from the image data itself, thus allowing retrospective gating (self-gating (SG)). Strategies exploit either changes of the k-space center signal (DC) 11 or use low-resolution images (Img-SG) reconstructed with high temporal resolution to extract motion-related gating signals 12-14. Img-SG has been applied successfully to 3D UTE lung imaging of human volunteers (3D-Img-SG, 15) and was demonstrated to perform superior to DC methods. The application to small animal imaging poses additional challenges, particularly the higher respiratory frequencies of small rodents. Thus, more refined methods for the extraction of a respiratory signal are needed. Retrospective gating further enables the reconstruction of images in different respiration stages from a single continuously acquired dataset. These can, for example, be used for the extraction of local ventilation, which can be derived from the difference in signal intensity in lung parenchyma between inspiration and expiration images. A similar approach has been used by Fischer et al. for human acquisitions 16, as well as by Bianchi et al., who recently demonstrated the applicability of ventilation maps derived from SG 2D UTE images to evaluate the functional impairment in a rat model of emphysema 17. Retrospective gating can further be applied to correct for bulk body motion during the acquisition, which often occurs albeit the animals are anesthetized 9. Bulk motion can be identified according to the Img-SG respiratory gating approach, but requires lower temporal resolution, as the interest is on identifying the changes in the animal position, not the characteristic of the movement itself. Eliminating bulk-motion corrupted data can be applied to further improve image fidelity 18. In this work, we investigated the feasibility of respiratory self-gated multistage reconstruction and bulk motion correction from 3D UTE in free-breathing rats using DC-SG and 3D-Img-SG, based on a 3D golden-angle radial sparse parallel (GRASP) imaging technique 14, 19. Signal-to-noise ratio (SNR) and normalized signal intensity (NSI) of lung parenchyma, image sharpness, and lung volume changes were studied to prove feasibility of the proposed method. METHODS Animal Handling Seven male Winstar rats (6 months, weight = 350 g ± 25 g) were imaged. Animal experiments were approved by the regional board for animal welfare of Tübingen, Germany, and conducted according to the German law for the welfare of animals and relevant regulations for care. All MRI images were acquired with a 7 Tesla small animal MRI system (Biospec 7/16, Bruker, Ettlingen, Germany) using a dedicated four-element rat thorax phased-array coil (RAPID Biomedical GmbH, Rimpar, Germany) with 48-mm inner diameter 20. After initiation of the anesthesia with 5% isoflurane in a mixture of N2:O2 (80:20), the rats were placed supine in the coil and the anesthesia gas was administered via a facial mask. During scanning, the isoflurane concentration was adapted to maintain a constant respiratory frequency of 60 to 70 cycles/min. The range of respiratory frequencies was recorded. Pulse Sequence A quasi-random acquisition scheme for 3D UTE 15, 21 was implemented on ParaVision 6.0 (Bruker, Ettlingen, Germany). Excitation was performed by applying a 4-μs block pulse, and the readout was started 6 µs after the end of the radiofrequency pulse, resulting in an effective TE of 8 μs. Trapezoid readout gradients with a ramp time of 116 µs and total readout time of 333 µs were used for the encoding of each spoke. Further sequence parameters were as follows: repetition time 2.4 ms; field of view = 7 × 7 × 7 cm³; flip angle = 3°; bandwidth = 300 kHz; matrix 200 × 200 × 200; six-fold oversampling yielding the acquisition of 752,736 projections; acquisition time 30 min. Image Reconstruction All images were reconstructed with in-house developed software implemented in MatLab (MathWorks, Natick, MA). The nonuniform k-space sampling density, introduced by the quasi-random profile ordering, was compensated using a Voronoi diagram 22. Two sets of time-resolved images were obtained from each acquisition. Full spatially resolved volume reconstructions with a temporal resolution of 30 s (12,500 spokes per frame) were used to identify bulk motion. Low spatially resolved images were used to retrieve the gating information. Here, reconstruction was performed on a 60 × 60 × 60 matrix from a reconstruction window spanning 80 spokes (192 ms), with an overlap of 40 spokes between subsequent frames, resulting in a total of 18,817 frames. This reconstruction was executed applying 3D-GRASP to minimize artifacts due to the high undersampling 13, 17. The regularization parameter λ was identified by preliminary analysis and chosen as λr = 10−6 in spatial domain and to λt = 10−3 along the time axes. The solver was terminated after 60 iterations in all cases. k-Space Center Self-Gating Signal Extraction For each acquisition, the DC-based SG signal was extracted from the magnitude of the k-space center signal and prefiltered with a Butterworth bandpass filter with a stopband attenuation of 30 dB between 0.5 to 2 Hz. For individual optimization of the filter characteristics, the DC-SG signal spectrum was estimated by calculating its power spectral density (PSD) with a discrete Fourier transform (FT) and was smoothed with a median filter of size 50. Lower and upper limits of the respiratory frequencies were identified as the mean value of the PSD between 0.5 and 0.6 Hz, plus 10 times its standard deviation (SD). The DC-SG signal was then filtered with the identified frequency limits by applying the same Butterworth bandpass filter. Filtered SG signals from all coil elements were combined by principal component analysis over the coil dimension 13. For further refinement, a spectrogram was calculated from the DC-SG signal prior to the individual second filter stage (short-time FT; window width 30,000; window overlap 29,900 points). The spectrogram allowed identification of changes of the respiratory peak frequency over time. When a shift was identified, a variable frequency DC-SG (VF-DC-G) was created. The SG signal was subdivided into 60 equally spaced windows. In each window, a PSD was calculated, and the respiratory frequency fc was identified as the maximum peak in the PSD. The SG-DC data within each window was filtered with a Butterworth bandpass filter between fc −0.1 and fc + 0.1 Hz. 3-D Image-Based SG Signal Extraction The 3D-Img-SG signal was retrieved from the sliding window images. For each image voxel, the magnitude values over time were extracted and the PSDs were estimated. Each PSD was smoothed with a median filter of dimension 300, and the values of the highest spectral peak between 0.5 and 2 Hz were identified. The voxel signal presenting the highest peak was chosen as SG signal and filtered, similar to the DC-SG and spline interpolated, to yield a SG signal for each spoke. Data Binning The SG signals were employed to separate the dataset into 10 bins, named stage 1 to stage 10 from the lowest diaphragm position to the highest (from inspiration (S1) to expiration (S10)). Each fraction of the signal between two consecutive peaks (end-expiration and end-inspiration) was divided into equally spaced bins, resulting into a continuous adaption of the threshold value depending on the peak amplitude of the gating signal. Bulk Motion Identification In order to automatically identify bulk movement from the temporal-resolved reconstruction, the central coronal slice was chosen and the 2D correlation coefficients were calculated for each time frame with respect to all others. It's assumed here that low correlation coefficients between frames indicate different positions of the animal. The most prevalent position can be selected as the frame presenting the highest mean values of the correlation coefficients. A practical threshold was set to reject data corresponding to time frames falling below the mean correlation. Lung Volumes and Image Sharpness Lung volumes were calculated for all stages to verify feasibility of the gating method. Voxels belonging to the lungs were segmented using a semiautomatic active contour method 23. Segmentations were manually checked to ensure accuracy. Lung volume changes ΔV were calculated as the lung volume at each stage minus the volume corresponding to maximum expiration (S10). The lung volumes were used to identify the respiratory position of the 10 stages. A threshold corresponding to one-tenth of the tidal volume above the minimal lung volume (expiration) was chosen, and any stage presenting a volume above this level was considered in inspiration. The bins representing expiration were combined, and the resulting reconstruction was compared with ungated images to investigate sharpness. Sharpness 24 was estimated as the average image gradient in each coronal slice, calculated via a 3 × 3 Sobel filter. Noise was removed from the gradient images by thresholding 25. These thresholds were determined from nongated images and held fixed for gated reconstructions of the same animal 14. Nonlung tissue was excluded by masking with the lung segmentations. Signal-to-Noise Ratio and Normalized Signal Intensity Calculation Signal-to-noise ratio and NSI in lung parenchyma were calculated for each acquisition in the posterior, medium, and anterior segment of the lung. Two regions of interest, one in lung parenchyma and one in muscle tissue, of approximately 150 pixels size were manually identified in each segment, excluding main vessels or bronchi. Signal-to-noise ratio was calculated as the mean signal in the parenchyma divided by the SD of the signal in the muscle, whereas NSI was calculated as the mean signal in the parenchyma divided by the mean signal in the muscle. Statistical Analysis Statistical analyses were performed with SPSS v20.0 (IBM Corp., Armonk, NY). Differences in ΔV, SNR, and NSI between different reconstructions were evaluated with an analysis of the variance test for repeated measures with Bonferroni's correction. Differences of the sharpness index between ungated and high-definition gated images were evaluated with a two-sided paired t test. Significance level was fixed at 0.05. Continuous values are presented as mean ± SD. RESULTS Figure 1 shows a representative example of the improvement in image quality obtained applying the GRASP reconstruction against simple gridding for the time-resolved reconstruction from which the 3D-Img-SG signal is extracted. The 3D-GRASP reconstruction yields sufficient image fidelity for deriving the 3D-Img-SG signal, whereas the gridding solution does not allow for recognition of any clear anatomical feature. The figure also indicates the pixel for which signal intensity variation over time determines the 3D-Img-SG signal. Figure 1Open in figure viewerPowerPoint Exemplification of the image quality obtained from the low-resolution, temporally resolved reconstruction used for the 3D-Img-SG. (a) Eight consecutive frames of a single coronal slice reconstructed with a gridding algorithm. The outline of the anatomical structure can be recognized only faintly. (b) Corresponding frames reconstructed with the 3D golden-angle radial sparse parallel protocol. Over the respiratory cycle, deviations from the expiratory position (red line) can be clearly appreciated. A black point in each frame in (b) indicates the pixel for which signal intensity variation over time determines the 3D-Img-SG signal. 3D-Img-SG; three-dimensional image-based self-gating. Figure 2 shows examples of different respiratory motion pattern. The DC-SG signal spectrum, the spectrogram, and resulting reconstructions for inspiration (S1) and expiration (S10) for the DC-SG and 3D-Img-SG are provided. For animal 1, the DC-SG signal presents a single clear spectral peak between 0.5 and 2 Hz, and the spectrogram shows a stable respiratory frequency over the acquisition. Both DC-SG and 3D-Img-DC achieve an effective gating and similar image quality. For animal 2, the spectrum does not indicate the presence of a clear respiratory peak, and the spectrogram does not identify any clear variation of respiratory rate over time. The DC-SG reconstruction does not identify the inspiratory position, and both S1 and S10 result blurred. With the 3D-Img-SG approach, inspiration and expiration images can be reconstructed with acceptable image quality. For animal 3, the spectrum presents two peaks (one lower and one higher than 1 Hz), and the spectrogram shows a decrease in respiration rate in the first part of the acquisition. A VF-DC-SG reconstruction was therefore applied, showing a clear improvement over the SG-DC approach but still inferior quality when compared to the 3D-Img-SG approach. Being the only method achieving gating in all cases, further analyses were performed only on the 3D-Img-DC reconstructions. Figure 2Open in figure viewerPowerPoint Comparison of the implemented algorithm for three different respiration patterns. The spectrum of the DC-SG signal, (a) its spectrogram (b) and inspiration (SG1) and expiration (S10) images, for the different gating approaches are provided. Animal 1 presents with nearly constant respiratory rate, and 3D-Img-SG as well as DC-SG yield good image quality. Animal 2 shows an arbitrarily varying respiratory rate over time and DC-SG is not effective, whereas 3D-Img-DC still allows multistage reconstruction. In animal 3, the spectrogram shows the presence of a respiratory frequency shift at the beginning of the acquisition, and the image quality of DC-SG can be improved by introducing a variable frequency filter (VF-DC-SG). 3D-Img-SG; three-dimensional image-based self-gating; DC-SG; k-space center self-gating, VF-DC-SG, variable frequency k-space center self-gating. The applicability of the 3D-Img-DC technique for dynamic reconstruction of the respiratory cycle is presented in Figure 3. The changes of the lung volume, bronchial tree, and pulmonary vasculature, as well as the shift of the diaphragm, can be clearly appreciated in the images reconstructed in the different motion states S1 to S10. Due to the unique respiration pattern of small rodents (rapid inspiration, long expiration), expiration images can be improved by combining the data of S5 to S10, yielding a clear improvement in SNR (expiration) but improved image sharpness when compared to the ungated reconstruction (ungated). Figure 3Open in figure viewerPowerPoint Comparison of the reconstruction of all respiratory stages, from S1 (full inspiration) to S10 (full expiration), or one coronal and one axial slice of the same acquisition. The red line marks the position of the diaphragm in expiratory position in all coronal slices. The expiration reconstruction combines the stages S5 to S10 to obtain a higher definition expiratory gated dataset, which is clearly sharper than the ungated reconstruction. Estimated lung volume, SNR, and NSI are presented in Figure 4. Lung volumes are presented as the difference between lung volumes at each stage minus the volume of maximum inspiration (stage 10). The estimated lung volumes change in between different stages (P < 0.0001), decfsreasing from inspiration (7.02 ± 0.81 mL) to expiration (5.11 ± 0.76 mL), resulting in a tidal volume of 1.90 ± 0.30 mL. Both the estimated SNR and NSI are significantly different among the different stages (P < 0.0001), having the lowest values at inspiration and showing a trend to continuously increase toward expiration, until they reach a plateau around stage 5. Figure 4Open in figure viewerPowerPoint Mean and standard deviation of ΔV (a), SNR (b), and NSI (c). Results of multiple comparisons are also reported on the graphics above each bar, either as the number of stages that are significantly different with respect to that stage (P < 0.5) or as an asterisk when all other stages were significantly different. SNR, signal-to-noise ratio; NSI, normalized signal intensity. Lung volume changes among the respiratory phases indicate that, in six of the seven acquisitions, the expiration stage is represented by 50% of the acquisition, and for the remaining animal by 60%. These values were used to reconstruct the high-definition expiration image, which were then compared to the ungated reconstruction in terms of sharpness. A comparison between expiratory gated and ungated images, along with the corresponding gradient images, is presented as Supporting Figure S1. Ungated images appear free of motion artifact but show a clearly reduced image sharpness. The sharpness index resulted in 0.0037 ± 0.001 for ungated and 0.0056 ± 0.002 for expiratory gated images, corresponding to an increment of 32.51% ± 8.1%. This difference was statistically significant in each dataset (P < 0.001). Bulk motion was identified in four of the seven datasets. Figure 5 shows a representative example where severe motion was observed during the first half of the acquisition (Fig. 5a). The image quality improvement obtained restricting the reconstruction to data above the mean value of the correlation coefficient is shown. Compared to the expiration (S5−S10) reconstruction from the full data (Fig. 5b), the bulk motion-compensated reconstruction (Fig. 5c) yields clear improvements in image sharpness. The image reconstructed from data below the chosen threshold shows a different animal position. Figure 5Open in figure viewerPowerPoint Bulk movement recognition and removal. Plot (a) represents the correlation coefficients used to identify the presence of movement across reconstructed frames. Exemplarily, the mean value of the correlation coefficient was chosen as threshold (a, red line). The full acquisition (expiratory gated, S5−S10) is shown in (b). The black arrow points to a vessel whose sharpness increases in the reconstruction, excluding bulk motion (data above threshold, c). Images reconstructed from the residual data reveal a different motion state as clearly visible, for example, by the course of the bronchi pointed by the red arrow (d). DISCUSSION In this work, we have shown the feasibility of reconstructing images corresponding to different respiratory stages from 3D UTE acquisition of freely breathing rats using a quasi-random acquisition scheme paired with image-based self-gating with GRASP reconstruction. The time-resolved 3D images reconstructed with a temporal resolution of ∼190 ms allowed for the extraction of a 3D-Img-SG signal with sufficient fidelity for the identification of different motion stages. Our findings indicate that the DC-SG signal achieves varying efficiency for respiratory gating. In this work, it resulted in good gating efficiency when the animal presented with a regular respiratory frequency, whereas the DC-SG was not able to identify the respiratory positions in datasets for which the respiration rate was not consistent. Using varying filter frequencies along the signal may be advantageous in case the respiratory rate drifts in the acquisition.k-space center SG is likely less efficient in 3D imaging than in 2D imaging. The method is based on changes of the total energy of the MR signal introduced by spins moving in and out of the field-of-view during respiration. In 2D UTE, prominent through-plane motion is present 9, but in 3D imaging much smaller portion of tissue moves in and out of the field-of-view during respiration, particularly when using a small field-of-view coil, reducing signal contribution from the abdominal area. It can be speculated that different coil dimensions, animal size, or positioning may improve the efficiency of respiratory DC-SG in 3D UTE. It is also possible that acquiring the DC signal separately with a fixed gradient direction may help to achieve a better signal but at the expense of increasing TR and ultimately the acquisition time. Three-dimensional Img-SG in 3D UTE has been previously applied in a clinical setting in healthy subjects 15. Its application to small animals requires higher temporal resolution due to the higher respiratory frequency. In addition, the lower number of coil elements that is available for small animal imaging, compared to clinical applications, limits the theoretically achievable acceleration factor. With the use of a 3D GRASP method, sufficient fidelity for the reproducible extraction of 3D-Img-SG signals was obtained. Iterative reconstruction methods may also be useful in a clinical setting with patients presenting with rapid and irregular breathing. The excellent SNR of lung parenchyma resulting with 3D-UTE causes the automatic extraction of the 3D-Img-SG signal from the time-resolved images to be more challenging because the lung–liver contrast decreases. A novel algorithm to extract the SG signal not aiming for identification of the lung–liver interface was introduced. Here, the signal intensity changes of voxels in the vicinity of the lung–liver interface were used. Those voxels cover a part of the lung during inspiration and part of the liver during expiration. Suitable voxels can be identified fully automatically as those showing the highest spectral peak around the respiratory frequencies. This approach performed well and has the clear advantage of being straightforward and not requiring any user interaction. One additional useful feature of the image-based method is the ability to visualize and identify bulk motion. Relatively low temporal resolutions are sufficient for identification of bulk motion, and GRASP reconstruction is not necessarily needed. Out of a huge variety of methods for motion identification, in this work, the correlation between frames was employed to identify bulk motion of the animal. Exclusion of motion corrupted data could be shown to further increase image sharpness. Further work may address the reconstruction of motion-compensated images, which can exploit the whole signal, registering the motion-corrupted sections. Because rodents under anesthesia breathe in a highly nonsinusoidal way, the proportion of the respiratory cycle that each animal spends in the expiratory position can be derived from the variation of lung volume over the reconstructed respiratory stages. In the acquired datasets, this value resulted in 50% to 60%, in agreement with 9. The stages representing the same position were pooled together to obtain high-definition gated images, with results significantly sharper than the ungated datasets. Overall, as in the 2D UTE case 9, respiratory gating may not be necessary to achieve good image quality for lung imaging with 3D UTE, but it is important when maximum image sharpness is required or functional analysis shall be performed. Signal-to-noise ratio and NSI in lung parenchyma significantly differ between the various respiratory stages, reflecting the density changes expected due to lung inflation. This change in signal intensity could be exploited to extract functional information on local ventilation, as in previous 2D studies 16. Three-dimensional UTE offers better SNR due to the extremely low TE and isotropic resolution, whereas the 2D method is restricted to thick coronal slices. CONCLUSION In conclusion, we have demonstrated the feasibility of respiratory gated reconstructions from 3D quasi-random UTE acquisitions in rodents, achieving high SNR and image quality in the inspiration as well as the expiration stage. The observed changes in signal intensity over the respiratory cycle may be considered as a first step toward calculation of 3D isotropic lung ventilation maps, which may provide information about functional properties of the lung in selected animal models of lung diseases. Supporting Information Additional supporting information may be found in the online version of this article. Filename Description mrm26463-sup-0001-suppinfof1.doc45.5 KB Fig. S1. Comparison of the image quality between expiratory gated (S5−S10, a), the ungated image (S1−S10, b) and the respective gradient images (c, d) after masking and thresholding for noise removal. Please note: The publisher is not responsible for the content or functionality of any supporting information supplied by the authors. Any queries (other than missing content) should be directed to the corresponding author for the article. REFERENCES 1 Wild J, Marshall H, Bock M, Schad L, Jakob P, Puderbach M, Molinari F, Van Beek E, Biederer J. MRI of the lung (1/3): methods. Insights Imaging 2012; 3: 345– 353. 2 Hatabu H, Alsop DC, Listerud J, Bonnet M, Gefter WB. T2* and proton density measurement of normal human lung parenchyma using submillisecond echo time gradient echo magnetic resonance imaging. Eur J Radiol 1999; 29: 245– 252. 3 Zurek M, Boyer L, Caramelle P, Boczkowski J, Crémillieux Y. Longitudinal and noninvasive assessment of emphysema evolution in a murine model using proton MRI. Magn Reson Med 2012; 68: 898– 904. 4 Bianchi A, Dufort S, Fortin P, Lux F, Raffard G, Tassali N, Tillement O, Coll J, Crémillieux Y. In vivo MRI for effective non-invasive detection and follow-up of an orthotopic mouse model of lung cancer. NMR Biomed 2014; 27: 971– 979. 5 Johnson KM, Fain SB, Schiebler ML, Nagle S. Optimized 3D ultrashort echo time pulmonary MRI. Magn Reson Med 2013; 70: 1241– 1250. 6 Jungraithmayr W, Chuck N, Frauenfelder T, Weder W, Boss A. MR imaging by using very short echo–time sequences after syngeneic lung transplantation in mice. Radiology 2012; 265: 753– 761. 7 Bianchi A, Tibiletti M, Kjørstad Å, Birk G, Schad LR, Stierstorfer B, Rasche V, Stiller D. Three-dimensional accurate detection of lung emphysema in rats using ultra-short and zero echo time MRI. NMR Biomed 2015; 28: 1471– 1479. 8 Glover G, Pauly J. Projection reconstruction techniques for reduction of motion effects in MRI. Magn Reson Med 1992; 28: 275– 289. 9 Tibiletti M, Kjørstad Å, Bianchi A, Schad LR, Stiller D, Rasche V. Multistage self-gated lung imaging in small rodents. Magn Reson Med 2016; 75: 2448– 2454. 10 Runge VM, Clanton JA, Partain CL, James AE Jr. Respiratory gating in magnetic resonance imaging at 0.5 tesla. Radiology 1984; 151: 521– 5310. 11 Ehman RL, Felmlee JP. Adaptive technique for high-definition MR imaging of moving structures. Radiology 1989; 173: 255– 263. 12 Zurek M, Bessaad A, Cieslar K, Crémillieux Y. Validation of simple and robust protocols for high-resolution lung proton MRI in mice. Magn Reson Med 2010; 64: 401– 407. 13 Larson AC, Kellman P, Arai A, Hirsch GA, McVeigh E, Li D, Simonetti OP. Preliminary investigation of respiratory self-gating for free-breathing segmented cine MRI. Magn Reson Med 2005; 53: 159– 168. 14 Paul J, Divkovic E, Wundrak S, Bernhardt P, Rottbauer W, Neumann H, Rasche V. High-resolution respiratory self-gated golden angle cardiac MRI: Comparison of self-gating methods in combination with k-t SPARSE SENSE. Magn Reson Med 2015; 73: 292– 298. 15 Tibiletti M, Paul J, Bianchi A, Wundrak S, Rottbauer W, Stiller D, Rasche V. Multistage three-dimensional UTE lung imaging by image-based self-gating. Magn Reson Med 2016; 75: 1324– 1332. 16 Fischer A, Weick S, Ritter CO, Beer M, Wirth C, Hebestreit H, Jakob PM, Hahn D, Bley T, Köstler H. SElf-gated Non-Contrast-Enhanced FUnctional lung imaging (SENCEFUL) using a quasi-random fast low-angle shot (FLASH) sequence and proton MRI. NMR Biomed 2014; 27: 907– 917. 17 Bianchi A, Tibiletti M, Kjørstad Å, Birk G, Schad LR, Stierstorfer B, Rasche V, Stiller D. Three-dimensional accurate detection of lung emphysema in rats using ultra-short and zero echo time MRI. NMR Biomed 2015; 28: 1471– 1479. 18 Higano NS, Hahn AD, Tkach J A, Cao X, Walkup, LL, Thomen RP, Merhar SL, Kingma PS, Fain SB, Woods JC. Retrospective respiratory self-gating and removal of bulk motion in pulmonary UTE MRI of neonates and adults. Magn Reson Med 2017; 77: 1284– 1295. 19 Feng L, Grimm R, Block KT, Chandarana H, Kim S, Xu J, Axel L, Sodickson DK, Otazo R. Golden-angle radial sparse parallel MRI: combination of compressed sensing, parallel imaging, and golden-angle radial sampling for fast and flexible dynamic volumetric MRI. Magn Reson Med 2014; 72: 707– 717. 20 Berthel, D, Neumaier, M, Schüler, D, Kaulisch, T, Lanz, T, Stiller, D. Optimization of a four channel phased array coil for rat lung imaging at 7 T. In Proceedings of the 22th Annual Meeting of ISMRM, Milan, Italy, 2014. 21 Weick, S, Ehses, P, Blaimer, M, Breuer, FA, Jakob, PM. Improved retrospective self-gated human lung imaging using a quasi random sampling scheme. In Proceedings of the 19th Annual Meeting of ISMRM, Montréal, Canada, 2011. 22 Rasche V, Proksa R, Sinkus R, Bornert P, Eggers H. Resampling of data between arbitrary grids using convolution interpolation. IEEE Trans Med Imaging 1999; 18: 385– 392. 23 Lankton S, Tannenbaum A. Localizing region-based active contours. ing, IEEE Trans Image Process 2008; 17: 2029– 2039. 24 McGee KP, Manduca A, Felmlee JP, Riederer SJ, Ehman RL. Image metric-based correction (autocorrection) of motion effects: analysis of image metrics. J Magn Reson Imaging 2000; 11: 174– 181. 25 Otsu N. A threshold selection method from gray-level histograms. Automatica 1975; 11: 23– 27. Citing Literature Volume78, Issue2August 2017Pages 739-745 FiguresReferencesRelatedInformation
Referência(s)