MORSE-PI -- Flexible and artefact-free image reconstruction for structural and functional QSM and other phase-critical imaging applications

arXiv:2606.21336 · physics.med-ph · Submitted 2026-06-19 · Read on arXiv

Listen

Radio episode about this paper

Transcript

Introduction to the show: ident: Physics. The day in general physics: instruments, methods and results across the field.

Kai: We're looking at the paper: "MORSE-PI -- Flexible and artefact-free image reconstruction for structural and functional QSM and other phase-critical imaging applications".

Mira: Here's the gist of the paper. Phase imaging applications such as QSM are highly sensitive to noise amplifications, phase singularities, and other artefacts, particularly in challenging scenarios such as ultra-high field (7T), under-sampled or single-echo acquisitions.

Kai: We present a novel image reconstruction method, MORSE-PI, designed to produce high-SNR, artefact-free, and singularity-free phase images for both structural (SWI, QSM) and functional (fQSM) phasebased brain imaging.

Mira: Purpose of the Method MORSE-PI is introduced to address the sensitivity of phase imaging applications like QSM to noise amplifications, phase singularities, and other artefacts in challenging scenarios.

Kai: The method aims to produce high-SNR, artefact-free, and singularity-free phase images for both structural (SWI, QSM) and functional (fQSM) phasebased brain imaging.

Paper summary: Kai: So we're looking at this paper called "MORSE-PI -- Flexible and artefact-free image reconstruction for structural and functional QSM and other phase-critical imaging applications".

Mira: This paper is addressing how sensitive phase imaging techniques are to noise amplification, phase singularities, and other artifacts, especially in tough spots like ultra-high field seven tesla or when you only get a single echo.

Lev: So the main goal of this work is to create a way to get high signal-to-noise ratio images that are free from those kinds of problems for both structural structural imaging like SWI and QSM, and functional ones called fQSM.

Kai: Exactly, they introduce this new method called MORSE-PI to tackle those sensitivity issues head on.

Mira: The core idea is that it produces phase images that are high signal-to-noise ratio, free of artifacts, and without those pesky singularities.

Lev: They claim this works for both structural and functional phase based brain imaging, which is a big deal because those applications really need clean phase information to be useful.

Kai: The method extends a previous approach called MORSE by adding something they call a Virtual Reference Coil or VRC.

Mira: This VRC isn't just some random thing; it's built as a linear combination of the coil sensitivity maps, and they enhance the correlations between those coil elements using the noise covariance matrix.

Lev: They say this Virtual Reference Coil ensures robust signal support across the entire brain and is used to correct phase offsets in those MORSE derived coil sensitivity estimates, which results in artefact-free, high signal-to-noise ratio phase.

Kai: What's interesting is they build this VRC from smooth MORSE derived coil sensitivities instead of using individual coil images for construction.

Mira: This smoothing step is what helps them enhance the intercoil correlation between the coils, which strengthens that Virtual Reference Coil idea.

Lev: So, if you were trying to run this on real hardware, you’d need those smooth sensitivities calculated first before you can build that VRC structure.

Paper summary: Kai: Compared to other methods they looked at, like GRAPPA with ASPIRE phase correction or ESPIRiT and GRAPPA combined methods, MORSE-PI shows better robustness against things like noise amplification and aliasing.

Mira: They also see improved reproducibility in structural imaging when they test it at both three tesla and seven tesla.

Lev: That improved reproducibility is important because when you're trying to build real hardware, you need results that are consistent across different field strengths, not just one perfect setup.

Kai: They specifically point out that unlike ESPIRiT or GRAPPA combined with adaptive coil combination methods, MORSE-PI yields singularity free phase maps.

Mira: That means they are tackling a specific problem in phase imaging where you can get those weird sharp spikes in the data, and MORSE-PI seems to handle that better than what those other combinations do.

Lev: From an engineering standpoint, getting singularity free phase maps is a big win because it simplifies downstream processing; you don't have to spend so much time trying to remove these specific artifacts before you can even start the main reconstruction.

Kai: So, this method is designed to make the whole pipeline smoother from a computational perspective.

Mira: And they show high signal-to-noise ratio reconstructions even in really tough scenarios, like single echo EPI acquisitions at seven tesla.

Lev: That single echo scenario is particularly hard because you lose a lot of information, so achieving high SNR there is quite impressive.

Kai: It seems like the paper shows that MORSE-PI provides high signal-to-noise ratio, fold over free and singularity free phase images across various acquisitions.

Mira: This flexibility means it works for single echo and multi echo structural gradient echo scans as well as functional EPI scans.

Lev: And the method doesn't require you to add an extra reference body coil or multi echo reference scans for this to work, which is a nice thing.

Kai: The method also works with a wide range of B zero map estimation techniques and background field removal methods, and even dipole inversion algorithms and denoising pipelines.

Paper summary: Mira: It's quite versatile when you put it into a larger framework of image processing tools.

Lev: That versatility means it could potentially be integrated into many existing MRI analysis software packages without needing a complete overhaul of the reconstruction pipeline.

Kai: So, the MORSE-PI paper is about developing this flexible reconstruction method that delivers high quality phase images for various acquisitions at different field strengths.

Mira: It’s specifically targeting those hard cases where noise and artifacts usually kill the image quality in QSM or fQSM.

Lev: The main implication for real hardware is that if you can implement this VRC construction smoothly, you get more reliable results from your expensive high-field scanners.

Kai: So it boils down to using a Virtual Reference Coil built cleverly with noise covariance information to correct phase errors and keep the signal strong across the brain.

Mira: It’s about making the reconstruction process less dependent on having perfect coil data or ideal acquisition settings for those tricky phase sensitive measurements.

Lev: The limitation they mention is that you still need that initial noise calibration scan to get all this going, which is something you have to do anyway in many protocols.

Kai: So, it's a method that offers high quality phase images without needing extra scans, as long as you do the initial calibration scan.

Mira: This means it’s a practical tool for researchers wanting better results from QSM and fQSM studies on high field systems.

Lev: It’s about getting more consistent and artifact free data from those challenging acquisitions.

Kai: So, to wrap up this MORSE-PI paper, it's about using a Virtual Reference Coil enhanced by noise covariance to produce singularity free phase images for structural and functional QSM.

Mira: It shows better robustness at both three tesla and seven tesla compared to GRAPPA with ASPIRE phase correction.

Lev: The paper points out that you still need that initial noise calibration scan for this reconstruction method to function correctly.

Kai: That’s the essence of MORSE-PI, focusing on creating a flexible and artefact free way to get phase images for QSM and fQSM.

Conclusion: Kai: So this paper is about MORSE-PI, which stands for something that lets you reconstruct images without those nasty phase artifacts in QSM and fQSM scans.

Mira: Yeah, the authors are Dymerska and others, looking at how to make those phase images really clean across different field strengths.

Lev: They’re aiming for high signal-to-noise ratio results that actually hold up when you take things to seven tesla or when you only get a single echo scan.

Kai: It seems like the title itself says it's about being flexible and artifact free, which is exactly what they claim this method achieves.

Mira: The authors are showing how their approach tackles those noise amplifications and phase singularities that usually ruin structural and functional scans.

Lev: What matters for us is that they demonstrate improved reproducibility in structural imaging at both three tesla and seven tesla, which is a tough comparison to make.

Kai: So what this means for us on the hardware side is that if you’re running an expensive scanner, this method could give you more consistent results without needing extra complex setups.

Mira: It suggests that by using a virtual reference coil built with noise information, we can get high quality phase images without having to rely on perfect coil data or extra reference scans.

Lev: The catch is they still need that initial noise calibration scan, which is something you usually have to do anyway in many protocols.

Kai: And that's the main thing they flag—it’s not a magic fix; it depends on having that calibration measurement first.

Mira: It lays out a solid framework for how we can handle phase-critical imaging more reliably, even in challenging scenarios like single echo EPI scans.

Lev: So it’s less about inventing a whole new physical principle and more about clever mathematical correction using the noise covariance matrix.

Kai: It’s basically an advanced way to correct the phase offsets by building this smart virtual reference coil that understands the noise correlations.

Mira: It really shows how important those underlying assumptions are, because if your initial assumptions about the noise model are off, you lose that high SNR.

Lev: This makes me think about how we need to ensure our error correction models account for those non-ideal noise environments in real hardware testing.

Kai: So it's a solid step forward in making these complex imaging techniques more usable for actual clinical and research settings at higher fields.

Barbara Dymerska, Oliver Josephs, Benjamin James, Vahid Malekian, Nadine N. Graedel, Martina F. Callaghan

University College London

physics.med-ph

Submitted: 2026-06-19

Updated: 2026-06-19

Comments: 33 pages including supplementary material

Journal ref: Dymerska, B., Josephs, O., James, B. et al. MORSE-PI: flexible and artefact-free image reconstruction for structural and functional QSM and other phase-critical imaging applications. Magn Reson Mater Phy (2026)

DOI: 10.1007/s10334-026-01434-8

Code: https://github.com/fil-physics/gadgetron-matlab

License: http://creativecommons.org/licenses/by/4.0/

Importance score: 89/100

The gist: Phase imaging applications such as QSM are highly sensitive to noise amplifications, phase singularities, and other artefacts…

Key concepts

Virtual Reference Coil (VRC)
The VRC is a mathematical construct created by combining coil sensitivity maps. Crucially, its elements are correlated using the noise covariance matrix to ensure robust signal support across the entire brain. This VRC helps correct phase offsets in coil sensitivity estimates, leading to cleaner, high-SNR phase images.
Noise Covariance Matrix
This matrix quantifies how noise varies across different MRI coils and locations. By using this information, MORSE-PI enhances the spatial correlation between neighboring coil sensitivities before calculating the VRC. This enhancement ensures the VRC has consistent signal support throughout the field of view.
Phase Singularities
Phase singularities are specific artifacts that appear in phase images, often caused by noise or reconstruction errors. MORSE-PI is specifically designed to eliminate these singularities, resulting in singularity-free phase maps. This is vital for accurate quantitative measurements like QSM.
Structural and Functional QSM (SWI/fQSM)
Quantitative Susceptibility Mapping (QSM) measures magnetic susceptibility in the brain tissue. MORSE-PI is applied to both structural (SWI, QSM) and functional (fQSM) phase-based brain imaging. The method provides the high-quality phase images necessary for these techniques to accurately map tissue properties.

Terminology

Summary

Phase imaging applications such as QSM are highly sensitive to noise amplifications, phase singularities, and other artefacts, particularly in challenging scenarios such as ultra-high field (7T), under-sampled or single-echo acquisitions. p. 1

Where the paper says it

Abstract: Purpose: Phase imaging applications such as QSM are highly sensitive to noise amplifications, phase singularities, and other artefacts, particularly in challenging scenarios such as ultra-high field (7T), under-sampled or single-echo acquisitions. We present a novel image reconstruction method, MORSE-PI, designed to produce high-SNR, artefact-free, and singularity-free phase images for both structural (SWI, QSM) and functional (fQSM) phasebased brain imaging.

We present a novel image reconstruction method, MORSE-PI, designed to produce high-SNR, artefact-free, and singularity-free phase images for both structural (SWI, QSM) and functional (fQSM) phasebased brain imaging. p. 1
Where the paper says it

Abstract: Purpose: Phase imaging applications such as QSM are highly sensitive to noise amplifications, phase singularities, and other artefacts, particularly in challenging scenarios such as ultra-high field (7T), under-sampled or single-echo acquisitions. We present a novel image reconstruction method, MORSE-PI, designed to produce high-SNR, artefact-free, and singularity-free phase images for both structural (SWI, QSM) and functional (fQSM) phasebased brain imaging.

Purpose of the Method

MORSE-PI is introduced to address the sensitivity of phase imaging applications like QSM to noise amplifications, phase singularities, and other artefacts in challenging scenarios. The method aims to produce high-SNR, artefact-free, and singularity-free phase images for both structural (SWI, QSM) and functional (fQSM) phasebased brain imaging. p. 1

Where the paper says it

Abstract: Purpose: Phase imaging applications such as QSM are highly sensitive to noise amplifications, phase singularities, and other artefacts, particularly in challenging scenarios such as ultra-high field (7T), under-sampled or single-echo acquisitions. We present a novel image reconstruction method, MORSE-PI, designed to produce high-SNR, artefact-free, and singularity-free phase images for both structural (SWI, QSM) and functional (fQSM) phasebased brain imaging.

p. 1
Where the paper says it

Abstract: Purpose: Phase imaging applications such as QSM are highly sensitive to noise amplifications, phase singularities, and other artefacts, particularly in challenging scenarios such as ultra-high field (7T), under-sampled or single-echo acquisitions. We present a novel image reconstruction method, MORSE-PI, designed to produce high-SNR, artefact-free, and singularity-free phase images for both structural (SWI, QSM) and functional (fQSM) phasebased brain imaging.

Core Reconstruction Mechanism

MORSE-PI extends the previous approach, MORSE, by introducing a Virtual Reference Coil (VRC). The VRC is constructed as a linear combination of coil sensitivity maps, with correlations enhanced between coil elements using the noise covariance matrix. p. 1

Where the paper says it

Methods: MORSE-PI extends our previous approach, MORSE, by introducing a Virtual Reference Coil (VRC). The VRC is constructed as a linear combination of coil sensitivity maps, with correlations enhanced between coil elements using the noise covariance matrix.

p. 1
Where the paper says it

Methods: MORSE-PI extends our previous approach, MORSE, by introducing a Virtual Reference Coil (VRC). The VRC is constructed as a linear combination of coil sensitivity maps, with correlations enhanced between coil elements using the noise covariance matrix.

This VRC ensures robust signal support across the entire brain and is used to correct phase offsets in the MORSE-derived coil sensitivity estimates, resulting in artefact-free, high SNR phase. p. 1
Where the paper says it

The VRC is constructed as a linear combination of coil sensitivity maps, with correlations enhanced between coil elements using the noise covariance matrix. Such a VRC ensures robust signal support across the entire brain and is used to correct phase offsets in the MORSE-derived coil sensitivity estimates, resulting in artefact-free, high SNR phase.

p. 1
Where the paper says it

The VRC is constructed as a linear combination of coil sensitivity maps, with correlations enhanced between coil elements using the noise covariance matrix. Such a VRC ensures robust signal support across the entire brain and is used to correct phase offsets in the MORSE-derived coil sensitivity estimates, resulting in artefact-free, high SNR phase.

To improve robustness, the VRC is constructed from smooth MORSEderived coil sensitivities, rather than from individual coil images, having enhanced the inter-coil correlation. p. 4
Where the paper says it

11 , the original VRC approach often leads to phase singularities, which can move around depending on the selection of the seed voxel or the weights used for the linear combination of the signal from the separate coils. To improve the robustness of the VRC approach in MORSE-PI, we construct the VRC from smooth MORSEderived coil sensitivities, rather than from individual coil images, having enhanced the inter-coil correlation,

Key Improvements Over Previous Methods

MORSE-PI offers several advantages over existing methods in phase imaging. Specifically, it demonstrates greater robustness to artefacts such as noise amplification and aliasing, and shows improved reproducibility in structural imaging at both 3T and 7T. p. 1

Where the paper says it

Such a VRC ensures robust signal support across the entire brain and is used to correct phase offsets in the MORSE-derived coil sensitivity estimates, resulting in artefact-free, high SNR phase. Results: Compared to GRAPPA with ASPIRE phase correction, MORSE-PI demonstrates greater robustness to artefacts such as noise amplification and aliasing, and shows improved reproducibility in structural imaging at both 3T and 7T.

Unlike ESPIRiT and GRAPPA combined with adaptive coil combination methods, MORSE-PI yields singularity-free phase maps. p. 1
Where the paper says it

Results: Compared to GRAPPA with ASPIRE phase correction, MORSE-PI demonstrates greater robustness to artefacts such as noise amplification and aliasing, and shows improved reproducibility in structural imaging at both 3T and 7T. Unlike ESPIRiT and GRAPPA combined with adaptive coil combination methods, MORSE-PI yields singularity-free phase maps.

Furthermore, it enables high-SNR reconstructions even for the most challenging scenarios, such as single-echo EPI at 7T. p. 1
Where the paper says it

Unlike ESPIRiT and GRAPPA combined with adaptive coil combination methods, MORSE-PI yields singularity-free phase maps. MORSE-PI enables high-SNR reconstructions even for the most challenging scenarios, such as single-echo EPI at 7T.

Performance and Validation

The method was evaluated across various field strengths and acquisition types. The comparison with alternative image reconstruction schemes showed that MORSE-PI produced singularity-free phase images and high-quality CLEAR-SWI and QSM reconstructions. p. 9

Where the paper says it

2 Structural imaging at 7T and 3T - comparison with alternative image reconstruction schemes MORSE-PI was compared with the original MORSE implementation, as well as methods commonly used for phase imaging: vendor-provided GRAPPA26 + Adaptive Combine27 (with and without SVD) and state-of-the-art phase imaging with GRAPPA26 + ASPIRE14 , and ESPIRiT 18 . The GRAPPA + Adaptive Combine (with and without SVD) and GRAPPA + ASPIRE images were reconstructed using their default image reconstruction chains (via the Siemens Image Calculation Environment) on the native scanner hardware via the “retrospective reconstruction” facility.

p. 9
Where the paper says it

2 Structural imaging at 7T and 3T - comparison with alternative image reconstruction schemes MORSE-PI was compared with the original MORSE implementation, as well as methods commonly used for phase imaging: vendor-provided GRAPPA26 + Adaptive Combine27 (with and without SVD) and state-of-the-art phase imaging with GRAPPA26 + ASPIRE14 , and ESPIRiT 18 . The GRAPPA + Adaptive Combine (with and without SVD) and GRAPPA + ASPIRE images were reconstructed using their default image reconstruction chains (via the Siemens Image Calculation Environment) on the native scanner hardware via the “retrospective reconstruction” facility.

Reproducibility assessments indicated that QSM results reconstructed using MORSE-PI exhibited reduced variability compared to those obtained with GRAPPA+ASPIRE, indicative of improved reproducibility. p. 1
Where the paper says it

Such a VRC ensures robust signal support across the entire brain and is used to correct phase offsets in the MORSE-derived coil sensitivity estimates, resulting in artefact-free, high SNR phase. Results: Compared to GRAPPA with ASPIRE phase correction, MORSE-PI demonstrates greater robustness to artefacts such as noise amplification and aliasing, and shows improved reproducibility in structural imaging at both 3T and 7T.

p. 1
Where the paper says it

Such a VRC ensures robust signal support across the entire brain and is used to correct phase offsets in the MORSE-derived coil sensitivity estimates, resulting in artefact-free, high SNR phase. Results: Compared to GRAPPA with ASPIRE phase correction, MORSE-PI demonstrates greater robustness to artefacts such as noise amplification and aliasing, and shows improved reproducibility in structural imaging at both 3T and 7T.

Conclusion and Applicability

MORSE-PI flexibly provides high SNR, fold-over-free and singularity-free phase images, as demonstrated for single-echo and multi-echo structural GRE and functional EPI scans. It is well-suited for phasebased imaging techniques such as structural and functional QSM. p. 1

Where the paper says it

Conclusion: MORSE-PI offers a flexible and computationally efficient solution for generating high-SNR, artefact- and singularity-free phase images in both single- and multi-echo GRE and EPI acquisitions. This makes it particularly well-suited for structural and functional QSM, as well as other phase-based MRI applications.

The method requires only a noise calibration scan, which is routinely acquired in many MRI protocols. p. 5
Where the paper says it

Using the noise covariance matrix to ensure good correlation between coil sensitivities, prior to VRC formation, provides good VRC magnitude and phase support, typically over the entire brain without phase singularities, as explained below. In many MRI applications, the noise covariance matrix is calculated from short noise measurements (without any excitation of MR signal) prior to running the imaging sequence, and used to decorrelate, or whiten, the complex signals from separate coils.

Code and Data Availability Statement

The open-source code, as well as scripts necessary to compile a Gadgetron image and create a Docker container encapsulating the MORSE-PI reconstruction, are available on https://github.com/fil-physics/gadgetron-matlab. p. 6

Where the paper says it

Methods The MORSE-PI image reconstruction scheme has been deployed for a diverse range of 3T and 7T functional and quantitative structural imaging studies at The Functional Imaging Laboratory (The FIL), University College London. The open-source code, along with scripts for creating a Docker container that encapsulates the MORSE-PI reconstruction within the Gadgetron framework19, is available on GitHub23 .

Example 3T and 7T multi-echo GRE raw k-space datasets are available on https://zenodo. p. 9
Where the paper says it

For all structural data, the MORSE and MORSE-PI reconstructions used: 𝑁𝑟𝑒𝑓 = 6, 𝑁𝑜𝑟𝑑𝑒𝑟 = 4, gaussian smoothing filter 𝑤 = 6 voxels, and global regularisation factor of 𝜆 = 3x10-4 , as deployed in various in vivo studies at The FIL17 . Structural images were acquired at 3T and 7T using an in-house, but openly available, 3D multi-echo spoiled gradient echo acquisition configured with either proton density (PD), T1 or magnetisation transfer (MT) weighting.

org/records/13737462. In-house developed GRE sequence for Siemens 3T Prisma system is https://xip. p. 8
Where the paper says it

Table 1 Sequence parameters for 3T and 7T GRE and EPI scans on which MORSE-PI reconstruction was evaluated. Protocol for Structural Imaging: 3D Gradient Echo Field [T] Resolution [mm3 ] Acceleration Reference data Matrix [HF X AP x RL] TR [ms] TE, 𝚫TE [ms] TA [min:s] 7 0.

uclb.com/product/mri-pulse-sequence-for-a-multi-echo-spoiled-gradient-echo. p. 9
Where the paper says it

For all structural data, the MORSE and MORSE-PI reconstructions used: 𝑁𝑟𝑒𝑓 = 6, 𝑁𝑜𝑟𝑑𝑒𝑟 = 4, gaussian smoothing filter 𝑤 = 6 voxels, and global regularisation factor of 𝜆 = 3x10-4 , as deployed in various in vivo studies at The FIL17 . Structural images were acquired at 3T and 7T using an in-house, but openly available, 3D multi-echo spoiled gradient echo acquisition configured with either proton density (PD), T1 or magnetisation transfer (MT) weighting.

Inhouse developed B1 sequence for Siemens 7T Terra system is https://xip. p. 7
Where the paper says it

The 3T data were acquired on a Siemens Healthineers Prismafit (software version VE11c) using a body coil for transmission and a 64-channel head and neck array for reception. The 7T data were obtained on a Siemens Healthineers Terra (software version VE12u-SP01) with a Nova Medical head coil featuring 8 transmit and 32 receive channels, operated in a quadrature-like transmit mode.

uclb.com/product/mri pulse sequence for b1. p. 5
Where the paper says it

Using the noise covariance matrix to ensure good correlation between coil sensitivities, prior to VRC formation, provides good VRC magnitude and phase support, typically over the entire brain without phase singularities, as explained below. In many MRI applications, the noise covariance matrix is calculated from short noise measurements (without any excitation of MR signal) prior to running the imaging sequence, and used to decorrelate, or whiten, the complex signals from separate coils.

Our CLEAR SWI and QSM processing pipelines for MORSE-PI are available on https://github. p. 9
Where the paper says it

Full details of all acquisitions are provided in Table 1. MR phase imaging is commonly used for SWI and QSM, we therefore prepared image processing pipelines for these contrasts and share them on GitHub28 .

com/fil-physics/MPM QSM. p. 6
Where the paper says it

Methods The MORSE-PI image reconstruction scheme has been deployed for a diverse range of 3T and 7T functional and quantitative structural imaging studies at The Functional Imaging Laboratory (The FIL), University College London. The open-source code, along with scripts for creating a Docker container that encapsulates the MORSE-PI reconstruction within the Gadgetron framework19, is available on GitHub23 .

References

Reichenbach JR, Barth M, Haacke EM, Klarhöfer M, Kaiser WA, Moser E. High-Resolution MR Venography at 3. p. 2

Where the paper says it

This contrast also underpins blood oxygenation level-dependent (BOLD) functional MRI, which detects changes in blood oxygenation via associated susceptibility variations4 . With the increasing availability of high-field MRI systems (≥3T), there has been rapid growth in high-resolution, phase-based imaging techniques sensitive to the magnetic properties of tissue, such as SusceptibilityWeighted Imaging (SWI)5 , Quantitative Susceptibility Mapping (QSM)6 , and functional QSM (fQSM)7 .

p. 2
Where the paper says it

This contrast also underpins blood oxygenation level-dependent (BOLD) functional MRI, which detects changes in blood oxygenation via associated susceptibility variations4 . With the increasing availability of high-field MRI systems (≥3T), there has been rapid growth in high-resolution, phase-based imaging techniques sensitive to the magnetic properties of tissue, such as SusceptibilityWeighted Imaging (SWI)5 , Quantitative Susceptibility Mapping (QSM)6 , and functional QSM (fQSM)7 .

0 Tesla. J Comput Assist Tomogr. 2000;24(6):949. Haacke EM, Cheng NYC, House MJ, et al. p. 9
Where the paper says it

For all structural data, the MORSE and MORSE-PI reconstructions used: 𝑁𝑟𝑒𝑓 = 6, 𝑁𝑜𝑟𝑑𝑒𝑟 = 4, gaussian smoothing filter 𝑤 = 6 voxels, and global regularisation factor of 𝜆 = 3x10-4 , as deployed in various in vivo studies at The FIL17 . Structural images were acquired at 3T and 7T using an in-house, but openly available, 3D multi-echo spoiled gradient echo acquisition configured with either proton density (PD), T1 or magnetisation transfer (MT) weighting.

Imaging iron stores in the brain using magnetic resonance imaging. p. 2
Where the paper says it

Introduction Magnetic Resonance Imaging (MRI) is highly sensitive to subtle variations in magnetic susceptibility across the human brain, particularly when using T2*-weighted acquisitions with gradient-echo (GRE) or echo-planar imaging (EPI) readouts, and at high magnetic field strengths. These techniques enable the study of cerebral vasculature1 as well as regional variations in tissue myelin and iron — two major contributors to magnetic susceptibility contrast2,3 .

Magn Reson Imaging. p. 0
Where the paper says it

MORSE-PI – Flexible and artefact-free image reconstruction for structural and functional QSM and other phase-critical imaging applications. Barbara Dymerska1 , Oliver Josephs1 , Benjamin James1 , Vahid Malekian1 , Nadine N.

2005;23(1):1-25 doi:10.1016/j.mri.2004.10.001. Lee J, Shmueli K, Kang BT, et al. The contribution of myelin to magnetic susceptibility-weighted contrasts in high-field MRI of the brain. p. 2
Where the paper says it

Introduction Magnetic Resonance Imaging (MRI) is highly sensitive to subtle variations in magnetic susceptibility across the human brain, particularly when using T2*-weighted acquisitions with gradient-echo (GRE) or echo-planar imaging (EPI) readouts, and at high magnetic field strengths. These techniques enable the study of cerebral vasculature1 as well as regional variations in tissue myelin and iron — two major contributors to magnetic susceptibility contrast2,3 .

NeuroImage. 2012;59(4):3967-3975 doi:10.1016/j.neuroimage.2011.10.076. Shmueli K, de Zwart JA, van Gelderen P, Li TQ, Dodd SJ, Duyn JH. p. 4
Where the paper says it

3 MORSE-PI with Virtual Reference Coil To remove ambiguous phase offsets and singularities we adopt the concept of a Virtual Reference Coil (VRC)20 with scalar phase matching at a seed voxel21,22 . In de Zwart et al.

Magnetic susceptibility mapping of brain tissue in vivo using MRI phase data. p. 2
Where the paper says it

This contrast also underpins blood oxygenation level-dependent (BOLD) functional MRI, which detects changes in blood oxygenation via associated susceptibility variations4 . With the increasing availability of high-field MRI systems (≥3T), there has been rapid growth in high-resolution, phase-based imaging techniques sensitive to the magnetic properties of tissue, such as SusceptibilityWeighted Imaging (SWI)5 , Quantitative Susceptibility Mapping (QSM)6 , and functional QSM (fQSM)7 .

Magn Reson Med. 2009;62(6):1510-1522 doi:10.1002/mrm.22135. Balla DZ, Sanchez-Panchuelo RM, Wharton SJ, et al. Functional quantitative susceptibility mapping (fQSM). p. 2
Where the paper says it

This contrast also underpins blood oxygenation level-dependent (BOLD) functional MRI, which detects changes in blood oxygenation via associated susceptibility variations4 . With the increasing availability of high-field MRI systems (≥3T), there has been rapid growth in high-resolution, phase-based imaging techniques sensitive to the magnetic properties of tissue, such as SusceptibilityWeighted Imaging (SWI)5 , Quantitative Susceptibility Mapping (QSM)6 , and functional QSM (fQSM)7 .

NeuroImage. 2014;100:112-124 doi:10. p. 8
Where the paper says it

Protocol for Functional Imaging: 3D EPI Field [T] Resolution [mm3 ] Acceleration Reference data Matrix [HF x AP x RL] TR [ms] TE [ms] TA [s/vol] 7 0. 8 None, 8 inplane segments None Whole brain 160 x 240 x240 100.

1016/j.neuroimage.2014.06.011. Liu C, Susceptibility tensor imaging. p. 2
Where the paper says it

The broader adoption of these methods has been facilitated by advances in acquisition efficiency, image reconstruction methods, and improved understanding of phase contrast mechanisms. It is now well established that susceptibility contrast not only reflects tissue composition but also microstructural orientation, as exploited by methods such as Susceptibility Tensor Imaging (STI)8 .

Magn Reson Med. 2010;63(6):1471-1477 doi:10.1002/mrm.22482. Shin HG, Lee J, Yun YH, et al. χ-separation: Magnetic susceptibility source separation toward iron and myelin mapping in the brain. p. 2
Where the paper says it

It is now well established that susceptibility contrast not only reflects tissue composition but also microstructural orientation, as exploited by methods such as Susceptibility Tensor Imaging (STI)8 . In parallel, emerging techniques like magnetic susceptibility source separation have provided increasing insight into sub-voxel distributions of key biological substrates such as iron and myelin9,10 .

NeuroImage. 2021;240:118371 doi:10. p. 8
Where the paper says it

Protocol for Functional Imaging: 3D EPI Field [T] Resolution [mm3 ] Acceleration Reference data Matrix [HF x AP x RL] TR [ms] TE [ms] TA [s/vol] 7 0. 8 None, 8 inplane segments None Whole brain 160 x 240 x240 100.

1016/j.neuroimage.2021.118371. Chen J, Gong NJ, Chaim KT, Otaduy MCG, Liu C. Decompose quantitative susceptibility mapping (QSM) to sub-voxel diamagnetic and paramagnetic components based on gradient-echo MRI data. p. 2
Where the paper says it

This contrast also underpins blood oxygenation level-dependent (BOLD) functional MRI, which detects changes in blood oxygenation via associated susceptibility variations4 . With the increasing availability of high-field MRI systems (≥3T), there has been rapid growth in high-resolution, phase-based imaging techniques sensitive to the magnetic properties of tissue, such as SusceptibilityWeighted Imaging (SWI)5 , Quantitative Susceptibility Mapping (QSM)6 , and functional QSM (fQSM)7 .

NeuroImage. 2021;242:118477 doi:10.1016/j.neuroimage.2021.118477. Dymerska B, Eckstein K, Bachrata B, et al. p. 0
Where the paper says it

Graedel1 , Martina F. Callaghan1 1The Functional Imaging Laboratory, Department of Imaging Neuroscience, UCL Queen Square Institute of Neurology, University College London, UK Running Title: MORSE-PI Corresponding author: Barbara Dymerska Department of Imaging Neuroscience, UCL Queen Square Institute of Neurology, University College London, London, WC1N 3AR, UK Email: b.

Phase unwrapping with a rapid opensource minimum spanning tree algorithm (ROMEO). p. 9
Where the paper says it

The QSM pipeline included phase unwrapping and field map calculation using ROMEO30 , hybrid masking using ROMEO30 and SPM31 segmentation32 , background field removal via PDF33 , and dipole inversion using STARQSM34 . We used CLEAR-SWI and ROMEO from the MriResearchTools package (v4.

Magn Reson Med. 2021;85(4):2294-2308 doi:10.1002/mrm.28563. Parker DL, Payne A, Todd N, Hadley JR. Phase reconstruction from multiple coil data using a virtual reference coil. p. 1
Where the paper says it

We present a novel image reconstruction method, MORSE-PI, designed to produce high-SNR, artefact-free, and singularity-free phase images for both structural (SWI, QSM) and functional (fQSM) phasebased brain imaging. Methods: MORSE-PI extends our previous approach, MORSE, by introducing a Virtual Reference Coil (VRC).

Magn Reson Med. 2014;72(2):563-569 doi:10.1002/mrm.24932. de Zwart JA, Ledden PJ, Kellman P, van Gelderen P, Duyn JH. p. 4
Where the paper says it

3 MORSE-PI with Virtual Reference Coil To remove ambiguous phase offsets and singularities we adopt the concept of a Virtual Reference Coil (VRC)20 with scalar phase matching at a seed voxel21,22 . In de Zwart et al.

Design of a SENSE-optimized highsensitivity MRI receive coil for brain imaging. p. 2
Where the paper says it

Several solutions exist that perform well for GRE-based phase imaging, particularly at 3T and lower field strengths, as reviewed in detail elsewhere11,13. However, the challenges increase at ultra-high fields (≥7T), where the lack of a body coil (often used as a phase reference in SENSE-based reconstructions) and the steeper coil sensitivity profiles complicate the generation of high-SNR, whole-brain phase images.

Magn Reson Med. 2002;47(6):1218-1227 doi:10.1002/mrm.10169. Hammond KE, Lupo JM, Xu D, et al. p. 4
Where the paper says it

In de Zwart et al. and Hammond et al.

Development of a robust method for generating 7. p. 2
Where the paper says it

These artefacts can propagate into the final combined phase image. We previously introduced MORSE15: a robust and computationally efficient method for coil sensitivity estimation and regularised SENSE16 image reconstruction.

0 T multichannel phase images of the brain with application to normal volunteers and patients with neurological diseases. p. 1
Where the paper says it

Abstract: Purpose: Phase imaging applications such as QSM are highly sensitive to noise amplifications, phase singularities, and other artefacts, particularly in challenging scenarios such as ultra-high field (7T), under-sampled or single-echo acquisitions. We present a novel image reconstruction method, MORSE-PI, designed to produce high-SNR, artefact-free, and singularity-free phase images for both structural (SWI, QSM) and functional (fQSM) phasebased brain imaging.

NeuroImage. 2008;39(4):1682-1692 doi:10.1016/j.neuroimage.2007.10.037. Eckstein K, Bachrata B, Hangel G, et al. Improved susceptibility weighted imaging at ultra-high field using bipolar multi-echo acquisition and optimized image processing: CLEAR-SWI. p. 2
Where the paper says it

With the increasing availability of high-field MRI systems (≥3T), there has been rapid growth in high-resolution, phase-based imaging techniques sensitive to the magnetic properties of tissue, such as SusceptibilityWeighted Imaging (SWI)5 , Quantitative Susceptibility Mapping (QSM)6 , and functional QSM (fQSM)7 . The broader adoption of these methods has been facilitated by advances in acquisition efficiency, image reconstruction methods, and improved understanding of phase contrast mechanisms.

NeuroImage. 2021;237:118175 doi:10.1016/j.neuroimage.2021.118175. Dymerska B, Poser BA, Barth M, Trattnig S, Robinson SD. p. 4
Where the paper says it

Following this scalar phase matching, a VRC can be formed by a linear combination of the phase signals of all the receiver coils20 . As discussed in the review paper by Robinson et al.

A method for the dynamic correction of B0-related distortions in single-echo EPI at 7T. p. 1
Where the paper says it

Unlike ESPIRiT and GRAPPA combined with adaptive coil combination methods, MORSE-PI yields singularity-free phase maps. MORSE-PI enables high-SNR reconstructions even for the most challenging scenarios, such as single-echo EPI at 7T.

NeuroImage. 2018;168:321-331 doi:10.1016/j.neuroimage.2016.07.009. Robinson SD, Bachrata B, Eckstein K, et al. p. 4
Where the paper says it

Following this scalar phase matching, a VRC can be formed by a linear combination of the phase signals of all the receiver coils20 . As discussed in the review paper by Robinson et al.

Improved dynamic distortion correction for fMRI using singleecho EPI and a readout-reversed first image (REFILL). p. 6
Where the paper says it

Below, we describe the implementation and summarize the performance evaluation carried out on representative in vivo data prior to deployment. 1 Gadgetron and MATLAB implementation Figure 1 visualises the computational steps of the MORSE-PI phase correction applied to the first order coil sensitivities calculated using MORSE.

Hum Brain Mapp. p. 1
Where the paper says it

Abstract: Purpose: Phase imaging applications such as QSM are highly sensitive to noise amplifications, phase singularities, and other artefacts, particularly in challenging scenarios such as ultra-high field (7T), under-sampled or single-echo acquisitions. We present a novel image reconstruction method, MORSE-PI, designed to produce high-SNR, artefact-free, and singularity-free phase images for both structural (SWI, QSM) and functional (fQSM) phasebased brain imaging.

2023;44(15):5095-5112 doi:10.1002/hbm.26440.

Supplementary material

Figure S1 Noise covariance matrices illustrating the effects of whitening or increasing cross-channel correlation in a 64-channel coil at 3T. Figure S2 Reproducibility assessment at 7T and 3T for GRAPPA+ASPIRE and MORSE-PI image reconstructions of PD-weighted and T1-weighted scans. p. 9

Where the paper says it

ROMEO uses weighted echo averaging for field map calculation36, which is deemed the optimal solution for QSM if the echo times are chosen appropriately13,37 . This solution can, however, lead to propagation of phase singularities present in MORSE, GRAPPA + Adaptive Combine, and ESPIRiT reconstructions to the final result.

Figure S4 Mean and temporal standard deviation (tSD) maps from MORSE-PI reconstruction of a 7T accelerated single-echo 3D EPI time series, processed using different QSM pipelines. p. 1
Where the paper says it

Unlike ESPIRiT and GRAPPA combined with adaptive coil combination methods, MORSE-PI yields singularity-free phase maps. MORSE-PI enables high-SNR reconstructions even for the most challenging scenarios, such as single-echo EPI at 7T.

Figure S5 Evaluation of SVD-based coil compression for phase singularity removal prior to image reconstruction. p. 6
Where the paper says it

Below, we describe the implementation and summarize the performance evaluation carried out on representative in vivo data prior to deployment. 1 Gadgetron and MATLAB implementation Figure 1 visualises the computational steps of the MORSE-PI phase correction applied to the first order coil sensitivities calculated using MORSE.

Figure S6 Example of MORSE-PI reconstruction of a 7T B1+ map using the Bloch-Siegert frequency shift method. p. 6
Where the paper says it

Below, we describe the implementation and summarize the performance evaluation carried out on representative in vivo data prior to deployment. 1 Gadgetron and MATLAB implementation Figure 1 visualises the computational steps of the MORSE-PI phase correction applied to the first order coil sensitivities calculated using MORSE.

Figure S7 MORSE-PI reconstructions of multi-echo gradient-echo images acquired in the liver (3 T) and knee (7 T). p. 9
Where the paper says it

For all structural data, the MORSE and MORSE-PI reconstructions used: 𝑁𝑟𝑒𝑓 = 6, 𝑁𝑜𝑟𝑑𝑒𝑟 = 4, gaussian smoothing filter 𝑤 = 6 voxels, and global regularisation factor of 𝜆 = 3x10-4 , as deployed in various in vivo studies at The FIL17 . Structural images were acquired at 3T and 7T using an in-house, but openly available, 3D multi-echo spoiled gradient echo acquisition configured with either proton density (PD), T1 or magnetisation transfer (MT) weighting.

Authors Contributions

Dymerska: Study conception and design, Code contributions, Acquisition design, Acquisition of data, Analysis and interpretation, Drafting of manuscript, Critical revision. Josephs: Code contributions, Analysis and interpretation. p. 2

Where the paper says it

In addition to magnetic susceptibility-induced phase perturbations (i. , ΔB0-related phase terms), each coil also encodes phase contributions from its own sensitivity profile, as well as system-related effects including frequency drifts, timing errors and eddy currents.

James: Code contributions. p. 2
Where the paper says it

In addition to magnetic susceptibility-induced phase perturbations (i. , ΔB0-related phase terms), each coil also encodes phase contributions from its own sensitivity profile, as well as system-related effects including frequency drifts, timing errors and eddy currents.

Malekian: Acquisition design, Critical revision. p. 0
Where the paper says it

MORSE-PI – Flexible and artefact-free image reconstruction for structural and functional QSM and other phase-critical imaging applications. Barbara Dymerska1 , Oliver Josephs1 , Benjamin James1 , Vahid Malekian1 , Nadine N.

Graedel: Acquisition design, Critical revision. p. 0
Where the paper says it

MORSE-PI – Flexible and artefact-free image reconstruction for structural and functional QSM and other phase-critical imaging applications. Barbara Dymerska1 , Oliver Josephs1 , Benjamin James1 , Vahid Malekian1 , Nadine N.

Callaghan: Study conception and design, Code contributions, Acquisition design, Acquisition of data, Critical revision. p. 7
Where the paper says it

Residual phase singularities in E remain only in regions lacking magnitude support for the respective coil (see B1), and therefore do not propagate to the final phase image. 2 In vivo data acquisition MORSE-PI was evaluated using data collected following ethical approval from local Research Ethics Committee and informed consent from study participants.

Compliance with ethical standars

Funding: This research was funded in whole, or in part, by the Wellcome Trust [203147/Z/16/Z and 226793/Z/22/Z].

Conflict of interest: The authors have no competing interests to declare that are relevant to the content of this article. p. 8

Where the paper says it

1 Impact of cross-coil correlation on VRC spatial support The strength of the correlation between coil sensitivities can be adjusted by varying 𝑛 in Equation (5). We tested values of 𝑛 = -1, 0, 1, 2, 3 to evaluate the VRC magnitude support across the regions of interest, namely the brain and cervical spinal cord in 3T and 7T structural scans.

Ethical approval: All procedures performed in studies involving human participants were in accordance with the ethical standards of the institutional and/or national research committee and with the 1964 Helsinki declaration and its later amendments or comparable ethical standards. p. 7

Where the paper says it

Residual phase singularities in E remain only in regions lacking magnitude support for the respective coil (see B1), and therefore do not propagate to the final phase image. 2 In vivo data acquisition MORSE-PI was evaluated using data collected following ethical approval from local Research Ethics Committee and informed consent from study participants.

Informed consent: Informed consent was obtained from all individual participants included in the study. p. 7

Where the paper says it

Residual phase singularities in E remain only in regions lacking magnitude support for the respective coil (see B1), and therefore do not propagate to the final phase image. 2 In vivo data acquisition MORSE-PI was evaluated using data collected following ethical approval from local Research Ethics Committee and informed consent from study participants.

  1. Reichenbach JR, Barth M, Haacke EM, Klarhöfer M, Kaiser WA, Moser E. High-Resolution MR Venography at 3.0 Tesla. J Comput Assist Tomogr. 2000;24(6):949.

  2. Li J, Lin H, Liu T, et al. Quantitative susceptibility mapping (QSM) minimizes interference from cellular pathology in R2 estimation of liver iron concentration. p. 2

    Where the paper says it

    This contrast also underpins blood oxygenation level-dependent (BOLD) functional MRI, which detects changes in blood oxygenation via associated susceptibility variations4 . With the increasing availability of high-field MRI systems (≥3T), there has been rapid growth in high-resolution, phase-based imaging techniques sensitive to the magnetic properties of tissue, such as SusceptibilityWeighted Imaging (SWI)5 , Quantitative Susceptibility Mapping (QSM)6 , and functional QSM (fQSM)7 .

    J Magn Reson Imaging. p. 0
    Where the paper says it

    MORSE-PI – Flexible and artefact-free image reconstruction for structural and functional QSM and other phase-critical imaging applications. Barbara Dymerska1 , Oliver Josephs1 , Benjamin James1 , Vahid Malekian1 , Nadine N.

    2018;48(4):1069-1079 doi:10.1002/jmri.26019.

  3. Wei H, Dibb R, Decker K, et al. Investigating magnetic susceptibility of human knee joint at 7 tesla. p. 2

    Where the paper says it

    Introduction Magnetic Resonance Imaging (MRI) is highly sensitive to subtle variations in magnetic susceptibility across the human brain, particularly when using T2*-weighted acquisitions with gradient-echo (GRE) or echo-planar imaging (EPI) readouts, and at high magnetic field strengths. These techniques enable the study of cerebral vasculature1 as well as regional variations in tissue myelin and iron — two major contributors to magnetic susceptibility contrast2,3 .

    Magn Reson Med. Published online January 1, 2017:n/a-n/a doi:10. p. 2
    Where the paper says it

    Denser array coils further exacerbate this issue by producing more spatially localised and variable sensitivity patterns. A well-established reconstruction method for 7T phase imaging is ASPIRE, which is a fast online (i.

    1002/mrm.26596.

Reference IDs

arXiv id 2606.21336

Page 1 pg1

Page 2 pg2

Page 3 pg3

Page 4 pg4

Page 5 pg5

Page 6 pg6

Page 7 pg7

Page 8 pg8

Page 9 pg9 (Note: Page numbers are based on the provided text structure, not the actual arXiv numbering)

Summary of MORSE-PI

MORSE-PI flexibly provides high SNR, fold-over-free and singularity-free phase images, as demonstrated for single-echo and multi-echo structural GRE and functional EPI scans. The method naturally lends itself to imaging techniques, such as structural and functional QSM, that necessitate high-quality phase images. p. 1

Where the paper says it

Abstract: Purpose: Phase imaging applications such as QSM are highly sensitive to noise amplifications, phase singularities, and other artefacts, particularly in challenging scenarios such as ultra-high field (7T), under-sampled or single-echo acquisitions. We present a novel image reconstruction method, MORSE-PI, designed to produce high-SNR, artefact-free, and singularity-free phase images for both structural (SWI, QSM) and functional (fQSM) phasebased brain imaging.

Pruning redundant reasoning tokens

MORSE-PI is introduced to address the sensitivity of phase imaging applications like QSM to noise amplifications, phase singularities, and other artefacts in challenging scenarios such as ultra-high field (7T), under-sampled or single-echo acquisitions. The method aims to produce high-SNR, artefact-free, and singularity-free phase images for both structural (SWI, QSM) and functional (fQSM) phasebased brain imaging.

How the training signal is built

MORSE-PI extends the previous approach, MORSE, by introducing a Virtual Reference Coil (VRC). The VRC is constructed as a linear combination of coil sensitivity maps, with correlations enhanced between coil elements using the noise covariance matrix. This VRC ensures robust signal support across the entire brain and is used to correct phase offsets in the MORSE-derived coil sensitivity estimates, resulting in artefact-free, high SNR phase. The VRC is constructed from smooth MORSEderived coil sensitivities, rather than from individual coil images, having enhanced the inter-coil correlation. p. 4

Where the paper says it

11 , the original VRC approach often leads to phase singularities, which can move around depending on the selection of the seed voxel or the weights used for the linear combination of the signal from the separate coils. To improve the robustness of the VRC approach in MORSE-PI, we construct the VRC from smooth MORSEderived coil sensitivities, rather than from individual coil images, having enhanced the inter-coil correlation,

Where the gains come from

The gains come from leveraging the noise covariance matrix to enhance the spatial correlation between neighbouring coil sensitivities prior to VRC calculation. This ensures that the VRC, calculated as the complex sum of coil sensitivities, has consistent support across the entire field of view. p. 7

Where the paper says it

The VRC sensitivity in step D is then computed as the complex sum of the correlated coil sensitivities, using scalar phase matching as described in Equation (6). This process ensures robust VRC support across the entire brain (see VRC magnitude image in D).

The level of spatial correlation is controlled by repeatedly applying the inverse of the pre-whitening matrix (i. p. 5
Where the paper says it

Therefore, having initially decorrelated the separate coil signals, i. , whitened the data, we subsequently use the lower triangular matrix to enhance the correlation between the first order MORSE-derived coil sensitivities: 𝒃̃𝑟,1 = 𝑳 𝑛𝒃̂ 𝑟,1 (5) where 𝒃̂ 𝑟,1 ∈ ℂ 𝑁coil is a vector with the original first order coil sensitivity estimates in physical coil space at a given spatial location 𝑟.

e. n = 1,2,3, … in Equation (5)). p. 4
Where the paper says it

𝑬̂𝑟 v = 𝑼𝑟𝑺𝑟𝑽𝑟 𝐻 = ∑𝜎𝑟,𝑘𝒖𝑟,𝑘𝒗𝑟,𝑘 𝐻 𝑁𝑟𝑒𝑓 𝑘=1 (3) The voxel-wise decomposition in Equation (3) produces 𝑁ref pairs, indexed by 𝑘, of right singular vectors 𝒗𝑟,𝑘 𝜖 ℂ 𝑁ref and left singular vectors 𝒖𝑟,𝑘 𝜖 ℂ 𝑁coil , with singular values, 𝜎𝑟,𝑘𝜖 ℝ+. A subset of left singular vectors, (𝑁order ≤ 𝑁ref) with singular values substantially greater than zero, is used as “higher-order” sensitivity estimates accounting for situations such as chemical shifts, rapidly varying sensitivities (e.

MORSE-PI demonstrates greater robustness to artefacts such as noise amplification and aliasing, and shows improved reproducibility in structural imaging at both 3T and 7T. The comparison with alternative image reconstruction schemes showed that MORSE-PI produced singularity-free phase images and high-quality CLEAR-SWI and QSM reconstructions. Reproducibility assessments indicated that QSM results reconstructed using MORSE-PI exhibited reduced variability compared to those obtained with GRAPPA+ASPIRE, indicative of improved reproducibility.

MORSE-PI is a highly flexible reconstruction method that delivers high-quality magnitude and phase images at both 3T and 7T for various acquisitions. It does not require additional reference body coil or multi-echo reference scans. p. 2

Where the paper says it

, on the MRI console) approach that typically produces high-quality phase images from multi-echo GRE data without requiring additional reference scans or a body coil14. However, ASPIRE requires at least two echoes, and its phase image quality strongly depends on the underlying reconstruction of individual coil images, which may be affected by artefacts from parallel imaging methods such as GRAPPA.

MORSE-PI is compatible with a wide range of B0 map estimation techniques, background field removal methods, dipole inversion algorithms, and denoising pipelines. p. 9
Where the paper says it

bias field removal) and Laplacian phase unwrapping, and two options adjusted to the resolution of our data: high-pass filter of size [2 2 0] voxels and Minimal Intensity Projection (MIP) across 14 axial slices (for depicting vessel networks). The QSM pipeline included phase unwrapping and field map calculation using ROMEO30 , hybrid masking using ROMEO30 and SPM31 segmentation32 , background field removal via PDF33 , and dipole inversion using STARQSM34 .

Green — the paper says it, and the page is printed. Red — no passage carries it, so treat that sentence as unverified. The same check runs on your own text: check a summary against its source, $5.

Improvements for AI systems

  1. Virtual Reference Coil (VRC) construction enhancement: The system can construct a VRC from smooth MORSE-derived coil sensitivities, rather than from individual coil images to ensure spatial homogeneity of the VRC and avoid regions without support like ear canals or large vessels.

  2. Cross-coil correlation control: The AI system can leverage the noise covariance matrix to enhance spatial correlation between neighboring coil sensitivities by applying the inverse of the pre-whitening matrix, controlled by a power n (n=1 being sufficient), to ensure VRC magnitude and phase support across the entire brain and cervical spinal cord.

  3. Singularity-free phase output for challenging acquisitions: The improved system can generate singularity-free images from single-echo 3D EPI at 7T, which currently requires additional reference scans using existing methods.

  4. Superior QSM reconstruction quality: The system can produce high SNR, fold-over-free and singularity-free phase images for structural and functional QSM applications at both 3T and 7T, outperforming GRAPPA + ASPIRE in QSM due to lower noise floors.

  5. Robustness against acquisition artifacts: The reconstruction is shown to be unaffected by local noise amplification observed in GRAPPA reconstructions, leading to substantially lower noise floor across the brain and higher quality CLEAR-SWI and QSM outputs.

Abstract

Phase imaging applications such as QSM are highly sensitive to noise amplifications, phase singularities, and other artefacts, particularly in challenging scenarios such as ultra-high field (7T), under-sampled or single-echo acquisitions. We present a novel image reconstruction method, MORSE-PI, designed to produce high-SNR, artefact-free, and singularity-free phase images for both structural and functional phase-based brain imaging. MORSE-PI extends our previous approach, MORSE, by introducing a Virtual Reference Coil (VRC). The VRC is constructed as a linear combination of coil sensitivity maps, with correlations enhanced between coil elements using the noise covariance matrix. Such a VRC ensures robust signal support across the entire brain and is used to correct phase offsets in the MORSE-derived coil sensitivity estimates, resulting in artefact-free, high SNR phase. Compared to GRAPPA with ASPIRE phase correction, MORSE-PI demonstrates greater robustness to artefacts such as noise amplification and aliasing, and shows improved reproducibility in structural imaging at both 3T and 7T. Unlike ESPIRiT and GRAPPA combined with adaptive coil combination methods, MORSE-PI yields singularity-free phase maps. MORSE-PI enables high-SNR reconstructions even for the most challenging scenarios, such as single-echo EPI at 7T. Its efficient, containerised implementation using the Gadgetron framework supports deployment on the MRI scanner console during measurements. MORSE-PI offers a flexible and computationally efficient solution for generating high-SNR, artefact- and singularity-free phase images in both single- and multi-echo GRE and EPI acquisitions. This makes it particularly well-suited for structural and functional QSM, as well as other phase-based MRI applications. Its robustness and rapid computational time facilitate efficient deployment on scanners across field strengths.

Sources

Related papers