Improving accelerated MRI by deep learning with sparsified complex data

PURPOSE

To obtain high-quality accelerated MR images with complex-valued reconstruction from undersampled k-space data.

METHODS

The MRI scans from human subjects were retrospectively undersampled with a regular pattern using skipped phase encoding, leading to ghosts in zero-filling reconstruction. A complex difference transform along the phase-encoding direction was applied in image domain to yield sparsified complex-valued edge maps. These sparse edge maps were used to train a complex-valued U-type convolutional neural network (SCU-Net) for deghosting. A k-space inverse filtering was performed on the predicted deghosted complex edge maps from SCU-Net to obtain final complex images. The SCU-Net was compared with other algorithms including zero-filling, GRAPPA, RAKI, finite difference complex U-type convolutional neural network (FDCU-Net), and CU-Net, both qualitatively and quantitatively, using such metrics as structural similarity index, peak SNR, and normalized mean square error.

RESULTS

The SCU-Net was found to be effective in deghosting aliased edge maps even at high acceleration factors. High-quality complex images were obtained by performing an inverse filtering on deghosted edge maps. The SCU-Net compared favorably with other algorithms.

CONCLUSION

Using sparsified complex data, SCU-Net offers higher reconstruction quality for regularly undersampled k-space data. The proposed method is especially useful for phase-sensitive MRI applications.

Motion resilience of the balanced steady-state free precession geometric solution

PURPOSE

Many MRI sequences are sensitive to motion and its associated artifacts. The linearized geometric solution (LGS), a balanced steady-state free precession (bSSFP) off-resonance signal demodulation technique, is evaluated with respect to motion artifact resilience.

THEORY AND METHODS

The mechanism and extent of LGS motion artifact resilience is examined in simulated, flow phantom, and in vivo clinical imaging. Motion artifact correction capabilities are decoupled from susceptibility artifact correction when feasible to permit controlled analysis of motion artifact correction when comparing the LGS with standard and phase-cycle-averaged (complex sum) bSSFP imaging.

RESULTS

Simulations reveal that the LGS demonstrates motion artifact reduction capabilities similar to standard clinical bSSFP imaging techniques, with slightly greater resilience in high SNR regions and for shorter-duration motion. Flow phantom experiments assert that the LGS reduces shorter-duration motion artifact error by ∼24%-65% relative to the complex sum, whereas reconstructions exhibit similar error reduction for constant motion. In vivo analysis demonstrates that in the internal auditory canal/orbits, the LGS was deemed to have less artifact in 24%/49% and similar artifact in 76%/51% of radiological assessments relative to the complex sum, and the LGS had less artifact in 97%/81% and similar artifact in 3%/16% of assessments relative to standard bSSFP. Only 2 of 63 assessments deemed the LGS inferior to either complex sum or standard bSSFP in terms of artifact reduction.

CONCLUSION

The LGS provides sufficient bSSFP motion artifact resilience to permit robust elimination of susceptibility artifacts, inspiring its use in a wide variety of applications.

A data-driven T2 relaxation analysis approach for myelin water imaging: Spectrum analysis for multiple exponentials via experimental condition oriented simulation (SAME-ECOS)

PURPOSE

The decomposition of multi-exponential decay data into a T₂ spectrum poses substantial challenges for conventional fitting algorithms, including non-negative least squares (NNLS). Based on a combination of the resolution limit constraint and machine learning neural network algorithm, a data-driven and highly tailorable analysis method named spectrum analysis for multiple exponentials via experimental condition oriented simulation (SAME-ECOS) was proposed.

THEORY AND METHODS

The theory of SAME-ECOS was derived. Then, a paradigm was presented to demonstrate the SAME-ECOS workflow, consisting of a series of calculation, simulation, and model training operations. The performance of the trained SAME-ECOS model was evaluated using simulations and six in vivo brain datasets. The code is available at https://github.com/hanwencat/SAME-ECOS.

RESULTS

Using NNLS as the baseline, SAME-ECOS achieved over 15% higher overall cosine similarity scores in producing the T₂ spectrum, and more than 10% lower mean absolute error in calculating the myelin water fraction (MWF), as well as demonstrated better robustness to noise in the simulation tests. Applying to in vivo data, MWF from SAME-ECOS and NNLS was highly correlated among all study participants. However, a distinct separation of the myelin water peak and the intra/extra-cellular water peak was only observed in the mean T₂ spectra determined using SAME-ECOS. In terms of data processing speed, SAME-ECOS is approximately 30 times faster than NNLS, achieving a whole-brain analysis in 3 min.

CONCLUSION

Compared with NNLS, the SAME-ECOS method yields much more reliable T₂ spectra in a dramatically shorter time, increasing the feasibility of multi-component T₂ decay analysis in clinical settings.

Myelin water imaging data analysis in less than one minute

PURPOSE

Based on a deep learning neural network (NN) algorithm, a super fast and easy to implement data analysis method was proposed for myelin water imaging (MWI) to calculate the myelin water fraction (MWF).

METHODS

A NN was constructed and trained on MWI data acquired by a 32-echo 3D gradient and spin echo (GRASE) sequence. Ground truth labels were created by regularized non-negative least squares (NNLS) with stimulated echo corrections. Voxel-wise GRASE data from 5 brains (4 healthy, 1 multiple sclerosis (MS)) were used for NN training. The trained NN was tested on 2 healthy brains, 1 MS brain with segmented lesions, 1 healthy spinal cord, and 1 healthy brain acquired from a different scanner.

RESULTS

Production of whole brain MWF maps in approximately 33 ​s can be achieved by a trained NN without graphics card acceleration. For all testing regions, no visual differences between NN and NNLS MWF maps were observed, and no obvious regional biases were found. Quantitatively, all voxels exhibited excellent agreement between NN and NNLS (all R²>0.98, p ​< ​0.001, mean absolute error <0.01).

CONCLUSION

The time for accurate MWF calculation can be dramatically reduced to less than 1 ​min by the proposed NN, addressing one of the barriers facing future clinical feasibility of MWI.

Banding artifact removal for bSSFP imaging with an elliptical signal model

PURPOSE

Balanced steady-state free precession (bSSFP) imaging has broad clinical applications by virtue of its high time efficiency and desirable contrast. Unfortunately, banding artifact is often seen as a result of signal modulation due to B0 inhomogeneity. This study aims to develop an effective method for banding artifact suppression.

METHODS

bSSFP is analyzed with an elliptical signal model. A simple analytical "Geometric-Solution" (GS) is presented to demodulate the signal from B0 inhomogeneity dependence with phase-cycled bSSFP data from both a computer simulation and experiments using phantom and human subjects.

RESULTS

The proposed algorithm is able to remove banding artifacts completely. It also compares favorably with the complex sum (CS), which is considered one of the more efficient methods for banding artifact correction.

CONCLUSION

Using an elliptical signal model, an analytical solution to the bSSFP banding problem has been found and demonstrated with simulation as well as phantom and in vivo experiments.

Improving image quality for skipped phase encoding and edge deghosting (SPEED) by exploiting several sparsifying transforms

PURPOSE

To improve the image quality of skipped phase encoding and edge deghosting (SPEED) by exploiting several sparsifying transforms.

METHODS

The SPEED technique uses a skipped phase encoding (PE) step to accelerate MRI scan. Previously, a difference transform (DT) along PE direction is used to obtain sparse ghosted-edge maps, which were modeled by a double layer ghost model and was then deghosted by a least square error solution. In this work, it is hypothesized that enhanced sparsity, and thus improved image quality may be achievable with other sparsifying transforms, including discrete wavelet transform (DWT), discrete cosine transform (DCT), DWT combined with DT, and DCT combined with DT.

RESULTS

For images of human subjects, SPEED with DWT or DCT can yield higher image quality than DT only, especially for those images with low contrast. Reconstruction error can be further reduced if DWT or DCT are combined with DT.

CONCLUSION

Image sparsity can be enhanced with more advanced transforms, leading to higher reconstruction quality in SPEED imaging that is desirable for practical MRI applications.

Somatic mosaicism for the p.His1047Arg mutation in PIK3CA in a girl with mesenteric lipomatosis

We describe a patient who presented with a localized growth of mature fat tissue, which was surgically removed. MRI imaging identified diffuse increase in visceral adipose tissue. Targeted deep sequencing of the resected tissue uncovered a p.H1047R variant in PIK3CA, which was absent in blood. This report expands the phenotypic spectrum of mosaic PIK3CA mutations.

Accelerated MRI by SPEED with generalized sampling schemes

PURPOSE

To enhance the fast imaging technique of skipped phase encoding (PE) and edge deghosting (SPEED) for more general sampling options, and thus more flexibility in implementations and applications.

METHODS

SPEED uses skipped PE steps to accelerate MRI scan. Previously, the PE skip size was chosen from prime numbers only. This restriction has been relaxed in this study to allow choice of any integers rather than merely prime numbers. Various sampling patterns were studied under all possible combinations of PE skip size and PE shifts. A criterion based on the rank values of ghost phasor matrices was introduced to evaluate SPEED reconstruction.

RESULTS

The reconstruction quality was found to correlate with the rank value of the ghost phasor matrix and the skipped PE size N. A low-rank value indicates a singular matrix that causes failure of the SPEED reconstruction. Composite numbers combined with appropriately chosen PE shifts yielded satisfactory reconstruction results.

CONCLUSION

With properly chosen PE shifts, it was found that any integers, including both prime numbers and composite numbers, could be used as PE skip size for SPEED. This finding allows much more flexible data acquisition options that may lead to more freedom in practical implementations and applications.

Applications of stimulated echo correction to multicomponent T2 analysis

We propose a multicomponent fitting algorithm for multiecho T(2) data which allows for correction of T(2) distributions in the presence of stimulated echoes. Tracking the population of spins in many coherence pathways via the iterated method of the Extended Phase Graph algorithm allows for accurate quantification of echo magnitudes. The resulting decay curves allow for correction of errors due to nonideal refocusing pulses as a result of inhomogeneities in the B(1) transmit field. Non-Negative Least Squares fitting is used to quantify the magnitude of T(2) components at various T(2) values. This method, allowing calculation of the T(2) distribution with simultaneous extraction of the refocusing pulse flip angle, requires no change to image acquisition procedures and no extra data input. Validation by means of both simulations and in vivo data shows excellent interscan reproducibility while vastly improving the accuracy of extracted T(2) parameters in voxels where poor B(1) homogeneity leads to refocusing pulse flip angles significantly less than 180°. Most notably, myelin water fraction values in these regions are found to have increased consistency and accuracy.

Simplified skipped phase encoding and edge deghosting (SPEED) for imaging sparse objects with applications to MRA

The fast imaging method named skipped phase encoding and edge deghosting (SPEED) has been demonstrated to reduce scan time considerably with typical magnetic resonance imaging data. In this work, SPEED is simplified with improved efficiency to accelerate the scan of sparse objects; we refer to this method as S-SPEED. S-SPEED partially samples k-space into two interleaved data sets, each with the same skip size of N but a different relative shift in phase encoding. The sampled data are then Fourier transformed into two ghosted images with N aliasing ghosts. Given the sparseness of signal distribution, the ghosted images are simply modeled with a single-layer structure, analogous to that used in maximum-intensity projection. With an algorithm based on a least-square-error solution, a deghosted image is solved, and a residual map is output for quality control. S-SPEED can be generalized to include more layers with additional acquisitions for refined results. Without differential filtering and full central k-space sampling, S-SPEED reduces scan time further and achieves more straightforward reconstruction, as compared with SPEED. In this work, S-SPEED is applied to accelerate magnetic resonance angiography (MRA) by taking advantage of the sparse nature of MRA data. With sparse phantom data and in vivo phase contrast MRA data, S-SPEED is demonstrated to achieve satisfactory results with an acceleration factor of 5.5 using a single coil.

Correction for geometric distortion and N/2 ghosting in EPI by phase labeling for additional coordinate encoding (PLACE)

Echo-planar imaging (EPI) is vulnerable to geometric distortion and N/2 ghosting. These artifacts can be analyzed with an intuitive k-t space tool, and here we propose a simple method for their correction. In a slightly modified additional EPI acquisition, we sample the k-t space with a shift in k(y) by adding a small area to the phase-encoding (PE) gradient. Physically, the added gradient area creates a relative phase ramp across the object and directly encodes the undistorted original y-coordinate of each voxel into a phase difference between two distorted complex images, in a method called "phase labeling for additional coordinate encoding" (PLACE). The phase information is then used to map the mismapped signals back to their original locations for geometric and intensity correction. Smoothing of expanded complex data matrix effectively reduces noise in the differential phase map and allows subpixel warping. The two acquired images can also be averaged to effectively suppress the N/2 ghost. Efficient correction for both artifacts can be achieved with three acquisitions. These acquisitions can also serve as reference scans to correct for geometric distortion and/or N/2 ghost artifacts on all images in a time series. The technique was successfully demonstrated in phantom and animal studies.

Highly accelerated MRI by skipped phase encoding and edge deghosting with array coil enhancement (SPEED-ACE)

The fast MRI method of skipped phase encoding and edge deghosting (SPEED) is further developed with array coil enhancement, and thus is termed SPEED-ACE. In SPEED-ACE, k space is sparsely sampled with skipped phase encoding at every Nth step using a set of receiver coils simultaneously, similar to SENSE, leading to sensitivity-weighted images with up to N layers of overlapping aliasing ghosts. The ghosted images are edge enhanced by a differential filter to yield ghosted edge maps, in which the ghost overlapping layers are greatly reduced since the sparseness of edges reduces the chance of ghost overlapping. Typical ghosted edge maps can be adequately modeled with a double-layer structure. By using data from at least three coils through least-square-error minimization, a deghosted edge map is obtained and inverse-filtered into a final deghosted image. In this way, SPEED-ACE partially samples k space with a skip size of N by using multiple receiver coils in parallel, and obtains a fairly good deghosted image with an undersampling factor of N. SPEED-ACE is not limited to the double-layer ghost model, but can be generalized to include more layers of ghosts for more flexible and improved performance. As a new parallel imaging method, SPEED-ACE was tested using in vivo data to demonstrate the possibility of achieving undersampling factors even greater than the number of receiver coils, which is so far not achievable by other parallel imaging methods.

Two-point water-fat imaging with partially-opposed-phase (POP) acquisition: An asymmetric Dixon method

A novel two-point water-fat imaging method is introduced. In addition to the in-phase acquisition, water and fat magnetization vectors are sampled at partially-opposed-phase (POP) rather than exactly antiparallel as in the original Dixon method. This asymmetric sampling encodes more valuable phase information for identifying water and fat. From the magnitudes of the two complex images, a big and a small chemical component are first robustly obtained pixel by pixel and then used to form two possible error phasor candidates. The true error phasor is extracted from the two error phasor candidates through a simple procedure of regional iterative phasor extraction (RIPE). Finally, least-squares solutions of water and fat are obtained after the extracted error phasor is smoothed and removed from the complex images. For noise behavior, the effective number of signal averages NSA* is typically in the range of 1.87-1.96, very close to the maximum possible value of 2. Compared to earlier approaches, the proposed method is more efficient in data acquisition and straightforward in processing, and the final results are more robust. At both 1.5T and 0.3T, well separated and identified in vivo water and fat images covering a broad range of anatomical regions have been obtained, supporting the clinical utility of the method.

Nonlinear phase correction with an extended statistical algorithm

This paper presents a new magnetic resonance imaging (MRI) phase correction method. The linear phase correction method using autocorrelation proposed by Ahn and Cho (AC method) is extended to handle nonlinear terms, which are often important for polynomial expansion of phase variation in MRI. The polynomial coefficients are statistically determined from a cascade series of n-pixel-shift rotational differential fields (RDFs). The n-pixel-shift RDF represents local vector rotations of a complex field relative to itself after being shifted by n pixels. We have found that increasing the shift enhances the signal significantly and extends the AC method to handle higher order nonlinear phase error terms. The n-pixel-shift RDF can also be applied to improve other methods such as the weighted least squares phase unwrapping method proposed by Liang. The feasibility of the method has been demonstrated with two-dimensional (2-D) in vivo inversion-recovery MRI data.

Accelerating MRI by skipped phase encoding and edge deghosting (SPEED)

A fast imaging method called skipped phase encoding and edge deghosting (SPEED) is introduced. The k-space is sparsely sampled into three interleaved datasets, each with a skip-size N and a relative shift in phase encoding (PE). These datasets are separately reconstructed by 2DFT and edge-enhanced by a differential filter in the PE direction, resulting in edge maps with phase-shifted aliasing ghosts. The sparseness of edges reduces the chance of ghost overlapping. Typical ghosted-edge maps can be adequately modeled with only two dominating ghost layers that are resolved from a set of three equations using least-square error minimization, yielding N ghost maps of different orders that can be registered and averaged into a single deghosted-edge map for noise and artifact reduction. Finally, the deghosted-edge map is transformed into a deghosted image by an inverse filter. A few central k-space lines are collected without PE skip to aid the inverse filtering. SPEED has been demonstrated by in vivo data to reduce scan time considerably without noticeable artifacts. It has various potential applications, such as MR angiography (MRA), where the signal itself is sparse. As an independent method, SPEED can be combined with other fast imaging methods for further acceleration.

Linear combination of multiecho data: short T2 component selection

The myelin sheath, which is wrapped around the axons in the brain, can be affected by many diseases, resulting in cognitive and physical disability. Other work showed water in the myelin sheath has a T2 approximately 15 ms. The current standard technique to estimate the fraction of myelin water in vivo is to collect multiecho data and fit the decay curves using a nonnegative least-squares (NNLS) algorithm. A new algorithm was developed to calculate optimized coefficients which were used to linearly combine multiecho data to estimate the myelin water signal. A set of simulations showed the new technique was accurate over a broad range of myelin water signal. The myelin water fraction from brain regions in scans from five volunteers, estimated by the linear combination method, agreed with the myelin water fraction estimated by the standard technique. The strength of the new technique is that the linear combination does not assume an underlying T2 model and is 20,000 times faster than NNLS.

Quantitative evaluation of metal artifact reduction techniques

PURPOSE

To develop a technique to quantify artifact, and to use it to compare the effectiveness of several approaches to metal artifact reduction, including view angle tilting and increasing the slice select and image bandwidths (BWs), in terms of metal artifact reduction, noise, and blur.

MATERIALS AND METHODS

Nonmetallic replicas of two metal implants (stainless steel and titanium/chromium-cobalt femoral prostheses) were fabricated from wax, and MR images were obtained of each component immersed in water. The differences between the images of each metal prosthesis and its wax counterpart were measured. The contributions from noise and blur were isolated, resulting in a measure of the metal artifact. Several off-resonance artifact reduction techniques were assessed in terms of metal artifact reduction capability, as well as signal to noise ratio and blur.

RESULTS

Increasing the image BW from +/-16 kHz to +/-64 kHz was found to reduce the artifact by an average of 60%, while employing view angle tilting (VAT) alone was found to reduce the artifact by an average of 63%. The metal artifact reduction sequence (MARS), which combines several susceptibility artifact reduction techniques, resulted in the least amount of image distortion, reducing the artifact by an average of 79%.

CONCLUSION

The results indicate that while VAT alone (with an image BW of +/-16 kHz) resulted in the smallest amount of total energy and no reduction in the signal-to-noise ratio compared to a conventional spin-echo pulse sequence, MARS resulted in significantly less artifact and dramatically less blur.

Understanding phase maps in MRI: a new cutline phase unwrapping method

This paper describes phase maps. A review of the phase unwrapping problem is given. Different structures, in particular fringelines, cutlines, and poles, contained within a phase map are described and their origin and behavior investigated. The problem of phase unwrapping can then be addressed with a better understanding of the source of poles or inconsistencies. This understanding, along with some assumptions about what is being encoded in the phase of a magnetic resonance image, are used to derive a new method for phase unwrapping which relies only on the phase map. The method detects cutlines and distinguishes between noise-induced poles and signal undersampling poles based on the length of the fringelines. The method was shown to be robust to noise and successful in unwrapping challenging clinical cases.

Chemical shift imaging with spectrum modeling

A new chemical shift imaging technique was developed to efficiently obtain separate images for multiple chemical shift peaks from a set of spin-echo acquisitions. Information from localized NMR spectroscopy was used to model the chemical shift spectrum as sharp peaks with known resonance frequencies but unknown amplitudes. Based on this model, a set of spin-echo images with shifted 180 degrees RF pulses were acquired, in which the magnetization vectors of different chemical components were put into different directions. The amplitudes of the chemical shift peaks were obtained by solving nonlinear equations in a region-growing process. Experimental results on an ethanol phantom as well as a subject with silicone breast implants are presented. Magn Reson Med 46:126-130, 2001.

Quantitative assessment of an MR technique for reducing metal artifact: application to spin-echo imaging in a phantom

OBJECTIVE

To quantify image artifact reduction using a new technique (MARS--metal artifact reduction sequence) in vitro.

DESIGN

Coronal T1-weighted MR images were obtained through two metal phantoms (titanium/chromium-cobalt and stainless steel femoral prostheses) immersed in water. Comparison of artifact volume was made with images obtained using conventional and modified (MARS) T1-weighted sequences. Signal intensity values outside a range of +/-40% the average signal intensity for water were considered artifact and segmented into low or high signal artifact categories. Considering the arbitrary selection of this threshold value, volumetric calculations of artifact were also evaluated at +/-50%, 60%, 70%, and 80% the mean signal for water.

RESULTS

Conventional T1-weighted images produced 87% more low signal artifact and 212% more high signal artifact compared with the MARS modified T1-weighted images of the stainless steel prosthesis. Conventional T1-weighted images of the titanium prosthesis produced 84% more low signal artifact and 211% more high signal artifact than the MARS modified sequence. The level of artifact reduction was essentially uniform for the various threshold levels tested and was greatest at +/-20% the global signal intensity average for water.

CONCLUSION

The MARS technique reduces the volume of image signal artifact produced by stainless steel and titanium/chromium-cobalt femoral prostheses on T1-weighted spin-echo images in a tissue phantom model.

MRI of spinal hardware: comparison of conventional T1-weighted sequence with a new metal artifact reduction sequence

OBJECTIVE

This study was designed to compare diagnostic quality of MR images of patients with spinal hardware acquired using a conventional T1-weighted spin-echo sequence and a new metal artifact reduction sequence (MARS).

CONCLUSION

The new MARS sequence effectively reduces the degree of tissue-obscuring artifact produced by spinal fixation hardware and subjectively improves image quality compared with the conventional T1-weighted spin-echo sequence.

A fast implementation of the minimum spanning tree method for phase unwrapping

A new implementation of the minimum spanning tree (MST) phase unwrapping method is presented. The time complexity of the MST method is reduced from O(n2) to O(n log2 n), where n is the number of pixels in the phase map. Typical 256 x 256 phase maps from magnetic resonance imaging can be unwrapped in seconds, compared with tens of minutes with the O(n2) implementation. This makes the pixel-level MST method time efficient and practically attractive. Index Terms-Image processing, magnetic resonance imaging, medical imaging, phase unwrapping.

Metal artifact reduction sequence: early clinical applications

Artifact arising from metal hardware remains a significant problem in orthopedic magnetic resonance imaging. The metal artifact reduction sequence (MARS) reduces the size and intensity of susceptibility artifacts from magnetic field distortion. The sequence, which is based on view angle tilting in combination with increased gradient strength, can be conveniently used in conjunction with any spin-echo sequence and requires no additional imaging time. In patients with persistent pain after femoral neck fracture, the MARS technique allows visualization of marrow adjacent to hip screws, thus enabling diagnosis or exclusion of avascular necrosis. Other applications in the hip include assessment of periprosthetic soft tissues after hip joint replacement surgery, postoperative assessment after resection of bone tumors and reconstruction, and localization of unopacified methyl methacrylate cement prior to hip arthroplasty revision surgery. In the knee, the MARS technique allows visualization of structures adjacent to implanted metal staples, pins, or screws. The technique can significantly improve visualization of periprosthetic bone and soft-tissue structures even in patients who have undergone total knee arthroplasty. In patients with spinal fixation hardware, the MARS technique frequently allows visualization of the vertebral bodies and spinal canal contents. The technique can be helpful after wrist fusion or screw fixation of scaphoid fractures.

An analysis algorithm for measuring airway lumen and wall areas from high-resolution computed tomographic data

High-resolution computed tomography (HRCT) has been used to examine airway narrowing. We developed an automated computed tomographic image analysis algorithm (computed tomographic airway morphometry; CTAM) to measure airway lumen area (Ai ), airway wall area (Awa), and airway angle of orientation. Tubes of varying size were embedded in Styrofoam and then scanned at angles between 0 degrees and 50 degrees to assess the accuracy of measurements made with CTAM. Two excised pig lungs were fixed in inflation, sectioned, and scanned. Ai and Awa were measured planimetrically from the cut surfaces to optimize CTAM measurement parameters. In CTAM, Ai was defined according to an airway-size-dependent threshold value, and total Awa was determined through a score-guided erosion method. Results were compared with measurements made through a previously validated method (manual method). CTAM provided accurate measurements of the tubes' Ai values at all angles; Awa was overestimated in direct relation to airway size. The manual method underestimated Ai and overestimated Awa in a manner directly related to airway size as well as to airway angle of orientation. In the excised lung, the mean errors of Ai and Awa measurements made with CTAM were 0.52 +/- 0.24 mm(2) and 0.17 +/- 0.32 mm(2) (mean +/- SEM), respectively. Ai errors with the manual method were similar, but Awa was overestimated to a greater degree (6.3 +/- 0.38 mm(2); p < 0.01) and the error was proportional to Awa (r = 0.64; p < 0.01). CTAM allows accurate measurements of airway dimensions and angle of orientation.

Application of a three-point method for water-fat MR imaging in children

BACKGROUND

Effective fat suppression is desirable in clinical magnetic resonance imaging. Conventional frequency selective fat suppression is dependent on accurate prescan shimming and is subject to artifacts due to magnetic field inhomogeneity. Quadrature three-point water-fat imaging with direct phase encoding is an alternative technique for fat suppression that has been previously described in adult volunteers and patients.

OBJECTIVE

To evaluate the use of three-point water-fat imaging with direct phase encoding for fat-suppressed MR scans in children.

MATERIALS AND METHODS

Sixty-two three-point water-fat imaging studies were performed in 55 children 2 months to 18 years old. T 1-weighted fat-suppressed (water) images from this sequence were compared with frequency selective fat-suppressed images obtained in 15 patients. The reliability and subjective quality of the sequence were assessed in the remaining 47 cases.

RESULTS

High-quality fat suppression was achieved in all anatomic sites studied, even where frequency selective fat-suppression failed due to magnetic susceptibility artifact. The three-point water-fat sequence was visually preferred to the frequency selective fat saturation technique in 15/15 cases.

CONCLUSION

Three-point water-fat imaging has replaced the conventional frequency selective technique for fat suppression on T 1-weighted MR imaging at our institution.

Water-fat imaging with direct phase encoding

A new method is introduced for water-fat imaging. With three acquisitions, a general direct phase encoding (DPE) of the chemical shift information is achieved. Pixels containing both water and fat are solved directly. Pixels with only a single component are resolved with local and global orientation filters, which use phase information from neighboring pixels. The fact that a single component is more likely to be water than fat in living tissues is also useful. A second pass solution yields water and fat images with superior signal-to-noise ratio. Unlike other methods, DPE does not rely on the error-prone phase unwrapping; also, it easily handles disconnected tissues. Because the magnetization vectors of water and fat are sampled not only at parallel or antiparallel, they can be not only separated but also identified respectively, which is desirable for routine clinical work. DPE has been implemented on several imagers at various field strengths and has been demonstrated in a large number of clinical cases to be useful and robust in various parts of the body.

Inversion recovery image reconstruction with multiseed region-growing spin reversal

A new algorithm is introduced for inversion recovery (IR) image reconstruction. The original complex image is modeled as a product of three factors: magnitude, polarity, and a smoothly changing phase factor. The simple binary polarity factor is first unified by a region-growing spin reversal (RGSR) operation, allowing the phase factor to be extracted. Multiplying the complex conjugate of the phase factor with the original complex data yields the desired IR contrast. The RGSR process is repeated with multiple seeds distributed in the field of view (FOV), and the results are added together, enabling disconnected tissues in the FOV to be handled. The extracted phase factor is filtered to reduce noise and artifacts, without losing useful information. The method is fully automatic and has been used practically in a large number of clinical examinations. The algorithm may also be useful for phase correction in simple proton spectroscopic imaging.

Temporal phase unwrapping for CINE velocity imaging

A simple algorithm named temporal phase unwrapping (TPU) is introduced to address the phase aliasing problem in time-dependent phase contrast (CINE-PC) velocity imaging. The method exploits the temporal continuity of velocity field and unwraps the phase along time. TPU only involves a one-dimensional (1D) temporal integration; therefore, many complications in 2D or 3D spatial phase unwrapping are avoided. Differential velocity maps (DVM) between adjacent movie frames are first calculated from the complex MR images. The DVMs have no phase aliasing as the differential velocities are much smaller than the absolute velocities. Aliasing-free velocity maps are obtained by integrating the DVMs along the time direction provided an aliasing-free reference velocity map (RVM) is found as a starting point of the integration. Typically, such RVMs are always available within the cardiac cycle, especially in diastole where the blood flow is the lowest. In vivo results from fully automated processing and detailed discussion on noise behavior are presented.

A quantitative interpretation of IVIM measurements of vascular perfusion in the rat brain

Pulsed gradient spin echo (PGSE) sequences have been used to measure the signal loss of 19F in perfluorinated hydrocarbon blood substitutes moving within the vasculature of the rat brain in the experimental conditions of the study. The signal loss is not characterized by a single apparent pseudodiffusion coefficient. A simple vascular network model based on self-similarity has been used to calculate the shape of the signal loss. Excellent agreement with the experiment has been obtained showing that the IVIM measurements are sensitive to flow over a wide range of vessel diameters and flow rates. This model of vascular structure may serve well for other MR measurements that are sensitive to perfusion.

Two-point interference method for suppression of ghost artifacts due to motion

A two-point interference method is introduced for suppression of ghosting due to motion in magnetic resonance imaging. The method requires only two time-interleaved data acquisitions, without any monitoring of the motion. A postprocessing technique is used to produce a weighted sum of the two acquired images, in which ghosts are suppressed by interference through an automatic regional tuning procedure. The appropriate complex weighting factors are regionally chosen by minimizing the "gradient energy," which is defined as the sum of squared pixel values in the partial-derivative maps. The method was tested in both phantoms and volunteers with a variety of imaging protocols. The level of ghost suppression with the two-point method was found to be comparable to that of the three-point method described previously by the authors.

Ghost phase cancellation with phase-encoding gradient modulation

Motion artifacts are a dominant cause of magnetic resonance image quality degradation. Periodic or nearly periodic motion results in image replicates of the moving structures in spin-warp Fourier imaging. The replicates, or ghosts, propagate in the image in the phase-encoding, or y, direction. These ghosted images can be considered to consist of the time-averaged spin density I0 and a ghost mask g. A set of j ghosted images Ij may be acquired in which the ghost mask is intentionally phase shifted by varying amounts relative to I0 with interleaved acquisitions that have shifted phase-encoding orders or by acquiring multiple images during a single readout period in the presence of an oscillating phase-encoding gradient. The resulting complex images Ij have the same time-averaged spin density I0 but have ghost contributions gj that, on a pixel-by-pixel basis, trace part of a circle around I0. The source images Ij can then be used to estimate I0. Simulations and experiments with the phase-encoding gradient modulation method show good general ghost suppression for a variety of quasi-periodic motion sources including both respiratory-type artifacts and flow artifacts. The primary limitation of the method is the need for rapid gradient switching.

Quantitative interpretation of magnetization transfer

Magnetization transfer contrast (MTC) experiments using off-resonance irradiation have been performed with an agar gel model by systematically varying offset frequency, amplitude of the RF irradiation and gel concentration. The experimental results are shown to be quantitatively modelled by a two-pool system consisting of a liquid pool with a Lorentzian line shape and a small semisolid pool with a Gaussian lineshape. The fitted model yields physically realistic fundamental parameters with a T2 of the semisolid pool of 13 microseconds. Further analysis shows that the off-resonance irradiation MTC experiment had significant limitations in its ability to saturate the semisolid pool without directly affecting the liquid component.

K-space description for MR imaging of dynamic objects

The spatial frequency (k) space concept is extended to describe the imaging of time-dependent objects. This work builds on the existing k-space description of MRI and is useful for simplifying the analysis of explanations of motion artifacts, algorithms for the correction of motion, and efficient imaging schemes for dynamic objects. Specific examples of the use of this concept for different imaging techniques are presented.

Dynamic image reconstruction: MR movies from motion ghosts

It has been previously shown that an image with motion ghost artifacts can be decomposed into a ghost mask superimposed over a ghost-free image. The present study demonstrates that the ghost components carry useful dynamic information and should not be discarded. Specifically, ghosts of different orders indicate the intensity and phase of the corresponding harmonics contained in the quasi-periodically varying spin-density distribution. A summation of the ghosts weighted by appropriate temporal phase factors can give a time-dependent dynamic image that is a movie of the object motion. This dynamic image reconstruction technique does not necessarily require monitoring of the motion and thus is easy to implement and operate. It also has a shorter imaging time than point-by-point imaging of temporal variation, because the periodic motion is more efficiently sampled with a limited number of harmonics recorded in the motion ghosts. This technique was tested in both moving phantoms and volunteers. It is believed to be useful for dynamic imaging of time-varying anatomic structures, such as in the cardiovascular system.

Motion artifact reduction with three-point ghost phase cancellation

A novel method for "ghost" artifact suppression is introduced. It suppresses ghosts induced by motion in any direction, as well as other types of quasi-periodic signal modulation. Because it requires neither special hardware nor intensive data processing, it can be easily implemented on conventional magnetic resonance (MR) imagers. The method is based on the concept of decomposition of a ghosted complex image into a ghost mask and ideal image. A set of deliberately designed acquisitions are used to generate a set of ghosted complex images in which the ghost components are related in a simple manner. With use of equations describing image decomposition and ghost correlation, the ideal image can be calculated pixel by pixel. The ideal image obtained (representing the time-averaged spin-density distribution) is shown to be a truer representation of physical reality than the ghost-free image obtained with ordered phase encoding. In this technique, both interview and intraview effects are taken into account. The technique is also useful in simultaneously suppressing ghosts from multifrequency signal modulations such as respiratory and cardiac motions. The method was successfully tested with three time-interleaved, phase-encoding-order-shifted acquisitions. Experimental results have shown that it is a simple but effective technique.

Projection images of the position-velocity joint spin density distribution

A new concept of phase encoding called position-velocity combined oblique Fourier phase encoding is introduced. It encodes both spatial and velocity information in a single oblique direction in the position-velocity space. Using this method, two-dimensional projection images of the three-dimensional position-velocity joint spin density distribution along different directions can be obtained. These projection images can provide detailed information on the flow system under study. The imaging of the two-dimensional projection is less time consuming compared to three-dimensional Fourier flow imaging and can be easily implemented on a conventional magnetic resonance imaging scanner.

Differential flow imaging by NMR

A new method for spatially resolved NMR flow measurements, named differential flow imaging (DFI), is introduced and experimentally verified. The DFI technique is based on the fact that flow velocity in any direction may cause a pixel position shift in the phase-encoding direction of a 2DFT NMR image. In this method two flow-influenced magnitude images are obtained by properly encoding and/or compensating the flow velocity. A spatial map of the desired component of the flow velocity can consequently be calculated from these two images. Since the DFI technique uses only the magnitude information of the complex images, it is not sensitive to systematic phase errors in contrast to other methods which are based on the phase measurements. On the other hand, the DFI technique can be combined with the phase measurement methods to perform multidimensional flow measurements in a shorter data acquisition time when the phase errors are small or corrected.

Measurement of mean and variance of velocity fields within each voxel by NMR imaging

If a flow field exists within an object that is being imaged, there could be a velocity spectrum in each voxel due to the finite voxel size. Consequently, the mean value and the variance of velocity in each voxel can be defined and used to describe the velocity spectrum. An NMR (nuclear magnetic resonance) imaging technique for performing the spatially resolved measurement of the mean value and the variance of velocity as well as the conventional spin density is proposed. A theoretical formalism and experimental results are presented and show good agreement. This technique is expected to be useful in the spatially resolved measurement of random directional flow, such as the fluid flow in living tissue and through porous materials.

A fonnalism for generating multiparametric encoding gradients in NMR tomography

A general mathematical formalism for generating multiparametric NMR image encoding gradients is introduced. The new schematic approach enables one to construct any desired encoding gradient which may be used in an imaging sequence. Basic gradient waveforms which can be used as building blocks of the desired encoding gradients are presented. A matrix operator for obtaining the encoding gradient for any kind of phase encoding is derived. Specific examples illustrating how to obtain "pure" spatial, velocity, or acceleration encoding gradients for moving spins are presented.