Bochao Li
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Research Subjects
Research Interests
Bochao Li has a total of 80 co-authors. Here are the top co-authors by number of shared publications:
- Krishna Shrinivas Nayak
Ming Hsieh Department of Electrical and Computer Engineering Viterbi School of Engineering, University of Southern California Los Angeles California USA
5 shared publications - Kübra Keskin
Electrical and Computer Engineering University of Southern California Los Angeles California USA
5 shared publications - Nam Gyun Lee
Alfred E. Mann Department of Biomedical Engineering University of Southern California Los Angeles California USA
5 shared publications - Sophia X. Cui
Siemens Medical Solutions USA Los Angeles California USA
4 shared publications - Ye Tian
Ming Hsieh Department of Electrical and Computer Engineering, Viterbi School of Engineering University of Southern California Los Angeles California USA
4 shared publications - Ziwei Zhao
Ming Hsieh Department of Electrical and Computer Engineering, Viterbi School of Engineering University of Southern California Los Angeles California USA
4 shared publications
Latest Publications
- Open Access
- 20 June 2026
- Published by Wiley in Journal of Magnetic Resonance Imaging
- Vol. 64 (4) , 1175-1181
Imaging Near Spinal Fixation Hardware at 0. 55 T Compared With 3 T
- 20 June 2026
- Published by Wiley in Journal of Magnetic Resonance Imaging
- Vol. 64 (4) , 1175-1181
Open AccessImaging Near Spinal Fixation Hardware at 0. 55 T Compared With 3 T
- Open Access
- 22 December 2025
- Published by Springer Nature in Magnetic Resonance Materials in Physics, Biology and Medicine
- Vol. 39 (4) , 1-12
Distortion correction in TSE near titanium implants at 0.55 T using reversed frequency-encoding and model-based reconstruction
Objectives: To develop a method for imaging near titanium implants at 0.55 T, which enables the use of a low readout bandwidth for higher SNR while reducing in-plane geometric distortions. Materials and methods: A pair of turbo spin echo ( TSE ) data with opposite frequency-encoding directions is acquired. For each frequency direction, a gradient nonlinearity ( GNL )-corrected image is reconstructed with a model-based iterative reconstruction incorporating GNL. A susceptibility-induced displacement map along the readout direction is estimated from two GNL-corrected images. A single final image is reconstructed with the model-based reconstruction incorporating both GNL and metal-induced displacement fields using both k-space acquisitions. The proposed method is compared against TSE with view angle tilting ( VAT ) and slice encoding for metal artifact correction ( SEMAC ). Results: The proposed reconstruction method maintains spatial resolution compared to current correction techniques. Unlike image-domain correction, VAT, and SEMAC, it does not introduce spatial blurring at equivalent bandwidths. It is feasible to reduce distortion due to off-resonance at a low readout bandwidth, resulting in higher apparent SNR (1.4–1.6-fold) without blurring under current imaging settings. In phantoms, and in patients with total hip arthroplasty and spinal fusion, the proposed method provides clearer delineation of tissues compared to conventional methods. Discussion: The proposed GNL and off-resonance distortion correction method for imaging near metal at 0.55 T enables the use of low readout bandwidth, providing SNR improvements without the blurring typically associated with low-bandwidth VAT.
- 22 December 2025
- Published by Springer Nature in Magnetic Resonance Materials in Physics, Biology and Medicine
- Vol. 39 (4) , 1-12
Open AccessDistortion correction in TSE near titanium implants at 0.55 T using reversed frequency-encoding and model-based reconstruction
Objectives: To develop a method for imaging near titanium implants at 0.55 T, which enables the use of a low readout bandwidth for higher SNR while reducing in-plane geometric distortions. Materials and methods: A pair of turbo spin echo ( TSE ) data with opposite frequency-encoding directions is acquired. For each frequency direction, a gradient nonlinearity ( GNL )-corrected image is reconstructed with a model-based iterative reconstruction incorporating GNL. A susceptibility-induced displacement map along the readout direction is estimated from two GNL-corrected images. A single final image is reconstructed with the model-based reconstruction incorporating both GNL and metal-induced displacement fields using both k-space acquisitions. The proposed method is compared against TSE with view angle tilting ( VAT ) and slice encoding for metal artifact correction ( SEMAC ). Results: The proposed reconstruction method maintains spatial resolution compared to current correction techniques. Unlike image-domain correction, VAT, and SEMAC, it does not introduce spatial blurring at equivalent bandwidths. It is feasible to reduce distortion due to off-resonance at a low readout bandwidth, resulting in higher apparent SNR (1.4–1.6-fold) without blurring under current imaging settings. In phantoms, and in patients with total hip arthroplasty and spinal fusion, the proposed method provides clearer delineation of tissues compared to conventional methods. Discussion: The proposed GNL and off-resonance distortion correction method for imaging near metal at 0.55 T enables the use of low readout bandwidth, providing SNR improvements without the blurring typically associated with low-bandwidth VAT.
- Open Access
- 2 October 2025
- Published by Wiley in Magnetic Resonance in Medicine
- Vol. 95 (2) , 1098-1109
Acceleration of slice encoding for metal artifact correction at 0. 55 T using hexagonal sampling
Purpose: To evaluate a hexagonal sampling approach for accelerated slice encoding for metal artifact correction (SEMAC) at 0.55 T. Contemporary mid‐field systems (0.1 T–1.0 T) show tremendous potential for imaging near metal implants. However, the limited parallel‐imaging options necessitate alternative methods for scan time reduction. Methods: We apply retrospective hexagonal undersampling to current state‐of‐the‐art SEMAC with 2‐fold generalized autocalibrating partially parallel acquisitions–based parallel imaging at 0.55 T. The hexagonal sampling approach results in an additional 50% scan time reduction. Feasibility is evaluated with phantom experiments involving spinal fixation and total hip arthroplasty hardware, and in vivo experiments involving patients with spinal fusions with varying compositions and 1 volunteer with a total hip arthroplasty. Results: Hexagonal sampling provides an additional 50% scan time reduction with compatible image quality. The two tradeoffs are (i) a small increase in signal void due to the loss of signal from one SEMAC spectral bin during post‐acquisition filtering and (ii) position‐dependent signal‐to‐noise‐ratio reduction at locations close to the edge of the field of view in the phase‐encoding direction. Conclusion: We demonstrate that hexagonal sampling can provide 50% scan time reduction in addition to generalized autocalibrating partially parallel acquisitions/parallel imaging for SEMAC at 0.55 T without introducing substantial artifacts. This may be a valuable mechanism for reducing scan time at 0.55 T and other midfield strengths, where parallel‐imaging acceleration is limited.
- 2 October 2025
- Published by Wiley in Magnetic Resonance in Medicine
- Vol. 95 (2) , 1098-1109
Open AccessAcceleration of slice encoding for metal artifact correction at 0. 55 T using hexagonal sampling
Purpose: To evaluate a hexagonal sampling approach for accelerated slice encoding for metal artifact correction (SEMAC) at 0.55 T. Contemporary mid‐field systems (0.1 T–1.0 T) show tremendous potential for imaging near metal implants. However, the limited parallel‐imaging options necessitate alternative methods for scan time reduction. Methods: We apply retrospective hexagonal undersampling to current state‐of‐the‐art SEMAC with 2‐fold generalized autocalibrating partially parallel acquisitions–based parallel imaging at 0.55 T. The hexagonal sampling approach results in an additional 50% scan time reduction. Feasibility is evaluated with phantom experiments involving spinal fixation and total hip arthroplasty hardware, and in vivo experiments involving patients with spinal fusions with varying compositions and 1 volunteer with a total hip arthroplasty. Results: Hexagonal sampling provides an additional 50% scan time reduction with compatible image quality. The two tradeoffs are (i) a small increase in signal void due to the loss of signal from one SEMAC spectral bin during post‐acquisition filtering and (ii) position‐dependent signal‐to‐noise‐ratio reduction at locations close to the edge of the field of view in the phase‐encoding direction. Conclusion: We demonstrate that hexagonal sampling can provide 50% scan time reduction in addition to generalized autocalibrating partially parallel acquisitions/parallel imaging for SEMAC at 0.55 T without introducing substantial artifacts. This may be a valuable mechanism for reducing scan time at 0.55 T and other midfield strengths, where parallel‐imaging acceleration is limited.
- Open Access
- 17 September 2025
- Published by Wiley in Magnetic Resonance in Medicine
- Vol. 95 (1) , 430-441
Free‐breathing 3D pulmonary ventilation mapping at 0. 55 T using stack‐of‐spiral out‐in bSSFP
Purpose: To develop a free‐breathing pulmonary imaging technique that provides three‐dimensional ( 3D) structural images and regional ventilation maps, and to evaluate its repeatability and accuracy compared with two‐dimensional phase‐resolved functional lung ( PREFUL ) and global tidal volume. Methods: A free‐breathing 3D stack‐of‐spiral out‐in ( SOS out‐in) balanced steady‐state free precession (bSSFP) sequence with self‐navigators was designed to achieve 2‐mm isotropic resolution in 5 min. Respiratory‐resolved images were reconstructed using spatial L1 wavelet and temporal finite‐difference constraints. The 3D ventilation maps were generated based on Jacobian determinant of the estimated nonrigid deformations. Six healthy volunteers were scanned in supine and prone positions. Ventilation maps were compared with two‐dimensional PREFUL from two matched slices. Test–retest repeatability was assessed using Bland–Altman analysis. Correlations among the proposed method, PREFUL, and global tidal volume were evaluated. Results: In healthy volunteers, the SOS out‐in lung images provided sufficient vessel‐parenchyma contrast and boundary sharpness to support accurate ventilation estimation. Regional ventilation measurements from 3D SOS out‐in demonstrated good repeatability (relative differences < 10%). Ventilation maps from 3D SOS out‐in strongly correlated with PREFUL on a slice‐matched basis as well as with global tidal volume ( R 2 > 0.7, p < 0.001). Conclusion: The proposed method provides high‐quality respiratory‐resolved structural images and 3D ventilation mapping in a single 5‐min scan at 0. 55 T . Ventilation measurements are sensitive, consistent, and in good agreement with PREFUL and spirometry.
- 17 September 2025
- Published by Wiley in Magnetic Resonance in Medicine
- Vol. 95 (1) , 430-441
Open AccessFree‐breathing 3D pulmonary ventilation mapping at 0. 55 T using stack‐of‐spiral out‐in bSSFP
Purpose: To develop a free‐breathing pulmonary imaging technique that provides three‐dimensional ( 3D) structural images and regional ventilation maps, and to evaluate its repeatability and accuracy compared with two‐dimensional phase‐resolved functional lung ( PREFUL ) and global tidal volume. Methods: A free‐breathing 3D stack‐of‐spiral out‐in ( SOS out‐in) balanced steady‐state free precession (bSSFP) sequence with self‐navigators was designed to achieve 2‐mm isotropic resolution in 5 min. Respiratory‐resolved images were reconstructed using spatial L1 wavelet and temporal finite‐difference constraints. The 3D ventilation maps were generated based on Jacobian determinant of the estimated nonrigid deformations. Six healthy volunteers were scanned in supine and prone positions. Ventilation maps were compared with two‐dimensional PREFUL from two matched slices. Test–retest repeatability was assessed using Bland–Altman analysis. Correlations among the proposed method, PREFUL, and global tidal volume were evaluated. Results: In healthy volunteers, the SOS out‐in lung images provided sufficient vessel‐parenchyma contrast and boundary sharpness to support accurate ventilation estimation. Regional ventilation measurements from 3D SOS out‐in demonstrated good repeatability (relative differences < 10%). Ventilation maps from 3D SOS out‐in strongly correlated with PREFUL on a slice‐matched basis as well as with global tidal volume ( R 2 > 0.7, p < 0.001). Conclusion: The proposed method provides high‐quality respiratory‐resolved structural images and 3D ventilation mapping in a single 5‐min scan at 0. 55 T . Ventilation measurements are sensitive, consistent, and in good agreement with PREFUL and spirometry.
- Open Access
- 28 March 2025
- Published by Wiley in Magnetic Resonance in Medicine
- Vol. 94 (1) , 242-250
Imaging near titanium total hip arthroplasty at 0. 55 T compared with 3 T
Purpose: To compare 0.55 T and 3 T MRI for imaging patients with titanium total hip arthroplasty (THA). Patients with orthopedic metallic implants often require diagnostic imaging to evaluate adjacent tissues. MRI performance measures, including artifact levels and SNR, vary with field strength. Methods: Six patients with titanium THA were scanned with similar protocols at 0.55 T and 3 T, including proton density (PD) weighted turbo spin echo (TSE), PD TSE with view‐angle tilting (TSE + VAT), PD slice encoding for metal artifact correction (SEMAC), and short tau inversion recovery with SEMAC (STIR‐SEMAC). Images from both field strengths were scored by two readers and qualitatively and quantitatively compared. Results: Diagnostic confidence was significantly higher at 0.55 T compared to 3 T. Perceived metal artifact was substantially reduced at 0.55 T compared to 3 T. At 0.55 T, diagnostic imaging was achieved both without and with multi spectral imaging (MSI) for PD weighted images. Conclusion: Compared to 3 T, 0.55 T MRI offers substantially reduced metal artifacts and higher diagnostic confidence when imaging titanium THA. Advanced multi‐spectral techniques may not be required when the metallic components are entirely titanium.
- 28 March 2025
- Published by Wiley in Magnetic Resonance in Medicine
- Vol. 94 (1) , 242-250
Open AccessImaging near titanium total hip arthroplasty at 0. 55 T compared with 3 T
Purpose: To compare 0.55 T and 3 T MRI for imaging patients with titanium total hip arthroplasty (THA). Patients with orthopedic metallic implants often require diagnostic imaging to evaluate adjacent tissues. MRI performance measures, including artifact levels and SNR, vary with field strength. Methods: Six patients with titanium THA were scanned with similar protocols at 0.55 T and 3 T, including proton density (PD) weighted turbo spin echo (TSE), PD TSE with view‐angle tilting (TSE + VAT), PD slice encoding for metal artifact correction (SEMAC), and short tau inversion recovery with SEMAC (STIR‐SEMAC). Images from both field strengths were scored by two readers and qualitatively and quantitatively compared. Results: Diagnostic confidence was significantly higher at 0.55 T compared to 3 T. Perceived metal artifact was substantially reduced at 0.55 T compared to 3 T. At 0.55 T, diagnostic imaging was achieved both without and with multi spectral imaging (MSI) for PD weighted images. Conclusion: Compared to 3 T, 0.55 T MRI offers substantially reduced metal artifacts and higher diagnostic confidence when imaging titanium THA. Advanced multi‐spectral techniques may not be required when the metallic components are entirely titanium.
- Open Access
- 10 May 2024
- Published by Wiley in Magnetic Resonance in Medicine
- Vol. 92 (3) , 1115-1127
Repeat it without me: Crowdsourcing the T1 mapping common ground via the ISMRM reproducibility challenge
Purpose: T1 mapping is a widely used quantitative MRI technique, but its tissue‐specific values remain inconsistent across protocols, sites, and vendors. The ISMRM Reproducible Research and Quantitative MR study groups jointly launched a challenge to assess the reproducibility of a well‐established inversion‐recovery T1 mapping technique, using acquisition details from a seminal T1 mapping paper on a standardized phantom and in human brains.Methods: The challenge used the acquisition protocol from Barral et al. (2010). Researchers collected T1 mapping data on the ISMRM/NIST phantom and/or in human brains. Data submission, pipeline development, and analysis were conducted using open‐source platforms. Intersubmission and intrasubmission comparisons were performed.Results: Eighteen submissions (39 phantom and 56 human datasets) on scanners by three MRI vendors were collected at 3 T (except one, at 0.35 T). The mean coefficient of variation was 6.1% for intersubmission phantom measurements, and 2.9% for intrasubmission measurements. For humans, the intersubmission/intrasubmission coefficient of variation was 5.9/3.2% in the genu and 16/6.9% in the cortex. An interactive dashboard for data visualization was also developed: https://rrsg2020.dashboards.neurolibre.org.Conclusion: The T1 intersubmission variability was twice as high as the intrasubmission variability in both phantoms and human brains, indicating that the acquisition details in the original paper were insufficient to reproduce a quantitative MRI protocol. This study reports the inherent uncertainty in T1 measures across independent research groups, bringing us one step closer to a practical clinical baseline of T1 variations in vivo.
- 10 May 2024
- Published by Wiley in Magnetic Resonance in Medicine
- Vol. 92 (3) , 1115-1127
Open AccessRepeat it without me: Crowdsourcing the T1 mapping common ground via the ISMRM reproducibility challenge
Purpose: T1 mapping is a widely used quantitative MRI technique, but its tissue‐specific values remain inconsistent across protocols, sites, and vendors. The ISMRM Reproducible Research and Quantitative MR study groups jointly launched a challenge to assess the reproducibility of a well‐established inversion‐recovery T1 mapping technique, using acquisition details from a seminal T1 mapping paper on a standardized phantom and in human brains.Methods: The challenge used the acquisition protocol from Barral et al. (2010). Researchers collected T1 mapping data on the ISMRM/NIST phantom and/or in human brains. Data submission, pipeline development, and analysis were conducted using open‐source platforms. Intersubmission and intrasubmission comparisons were performed.Results: Eighteen submissions (39 phantom and 56 human datasets) on scanners by three MRI vendors were collected at 3 T (except one, at 0.35 T). The mean coefficient of variation was 6.1% for intersubmission phantom measurements, and 2.9% for intrasubmission measurements. For humans, the intersubmission/intrasubmission coefficient of variation was 5.9/3.2% in the genu and 16/6.9% in the cortex. An interactive dashboard for data visualization was also developed: https://rrsg2020.dashboards.neurolibre.org.Conclusion: The T1 intersubmission variability was twice as high as the intrasubmission variability in both phantoms and human brains, indicating that the acquisition details in the original paper were insufficient to reproduce a quantitative MRI protocol. This study reports the inherent uncertainty in T1 measures across independent research groups, bringing us one step closer to a practical clinical baseline of T1 variations in vivo.
- Open Access
- 1 February 2024
- Published by Polytechnique Montreal
- p. 23
Paper is not enough: Crowdsourcing the T1 mapping common ground via the ISMRM reproducibility challenge
- 1 February 2024
- Published by Polytechnique Montreal
- p. 23
Open AccessPaper is not enough: Crowdsourcing the T1 mapping common ground via the ISMRM reproducibility challenge
- Open Access
- 9 October 2023
- Published by Centre de Recherche de l'Institut Universitaire de Geriatrie de Montreal
- Vol. 3 (34) , 14
Results of the ISMRM 2020 joint Reproducible Research & Quantitative MR study groups reproducibility challenge on phantom and human brain T1 mapping
- 9 October 2023
- Published by Centre de Recherche de l'Institut Universitaire de Geriatrie de Montreal
- Vol. 3 (34) , 14
Open AccessResults of the ISMRM 2020 joint Reproducible Research & Quantitative MR study groups reproducibility challenge on phantom and human brain T1 mapping
- Open Access
- 20 November 2022
- Published by Wiley in Magnetic Resonance in Medicine
- Vol. 89 (4) , 1522-1530
Lung parenchyma transverse relaxation rates at 0.55 T
Purpose: To determine R2 and transverse relaxation rates in healthy lung parenchyma at 0.55 T. This is important in that it informs the design and optimization of new imaging methods for 0.55T lung MRI.Methods: Experiments were performed in 3 healthy adult volunteers on a prototype whole‐body 0.55T MRI, using a custom free‐breathing electrocardiogram‐triggered, single‐slice echo‐shifted multi‐echo spin echo (ES‐MCSE) pulse sequence with respiratory navigation. Transverse relaxation rates R2 and and off‐resonance ∆f were jointly estimated using nonlinear least‐squares estimation. These measurements were compared against R2 estimates from T2‐prepared balanced SSFP (T2‐Prep bSSFP) and estimates from multi‐echo gradient echo, which are used widely but prone to error due to different subvoxel weighting.Results: The mean R2 and values of lung parenchyma obtained from ES‐MCSE were 17.3 ± 0.7 Hz and 127.5 ± 16.4 Hz (T2 = 61.6 ± 1.7 ms; = 9.5 ms ± 1.6 ms), respectively. The off‐resonance estimates ranged from −60 to 30 Hz. The R2 from T2‐Prep bSSFP was 15.7 ± 1.7 Hz (T2 = 68.6 ± 8.6 ms) and from multi‐echo gradient echo was 131.2 ± 30.4 Hz ( = 8.0 ± 2.5 ms). Paired t‐test indicated that there is a significant difference between the proposed and reference methods (p < 0.05). The mean R2 estimate from T2‐Prep bSSFP was slightly smaller than that from ES‐MCSE, whereas the mean and estimates from ES‐MCSE and multi‐echo gradient echo were similar to each other across all subjects.Conclusions: Joint estimation of transverse relaxation rates and off‐resonance is feasible at 0.55 T with a free‐breathing electrocardiogram‐gated and navigator‐gated ES‐MCSE sequence. At 0.55 T, the mean R2 of 17.3 Hz is similar to the reported mean R2 of 16.7 Hz at 1.5 T, but the mean of 127.5 Hz is about 5–10 times smaller than that reported at 1.5 T.
- 20 November 2022
- Published by Wiley in Magnetic Resonance in Medicine
- Vol. 89 (4) , 1522-1530
Open AccessLung parenchyma transverse relaxation rates at 0.55 T
Purpose: To determine R2 and transverse relaxation rates in healthy lung parenchyma at 0.55 T. This is important in that it informs the design and optimization of new imaging methods for 0.55T lung MRI.Methods: Experiments were performed in 3 healthy adult volunteers on a prototype whole‐body 0.55T MRI, using a custom free‐breathing electrocardiogram‐triggered, single‐slice echo‐shifted multi‐echo spin echo (ES‐MCSE) pulse sequence with respiratory navigation. Transverse relaxation rates R2 and and off‐resonance ∆f were jointly estimated using nonlinear least‐squares estimation. These measurements were compared against R2 estimates from T2‐prepared balanced SSFP (T2‐Prep bSSFP) and estimates from multi‐echo gradient echo, which are used widely but prone to error due to different subvoxel weighting.Results: The mean R2 and values of lung parenchyma obtained from ES‐MCSE were 17.3 ± 0.7 Hz and 127.5 ± 16.4 Hz (T2 = 61.6 ± 1.7 ms; = 9.5 ms ± 1.6 ms), respectively. The off‐resonance estimates ranged from −60 to 30 Hz. The R2 from T2‐Prep bSSFP was 15.7 ± 1.7 Hz (T2 = 68.6 ± 8.6 ms) and from multi‐echo gradient echo was 131.2 ± 30.4 Hz ( = 8.0 ± 2.5 ms). Paired t‐test indicated that there is a significant difference between the proposed and reference methods (p < 0.05). The mean R2 estimate from T2‐Prep bSSFP was slightly smaller than that from ES‐MCSE, whereas the mean and estimates from ES‐MCSE and multi‐echo gradient echo were similar to each other across all subjects.Conclusions: Joint estimation of transverse relaxation rates and off‐resonance is feasible at 0.55 T with a free‐breathing electrocardiogram‐gated and navigator‐gated ES‐MCSE sequence. At 0.55 T, the mean R2 of 17.3 Hz is similar to the reported mean R2 of 16.7 Hz at 1.5 T, but the mean of 127.5 Hz is about 5–10 times smaller than that reported at 1.5 T.
- Open Access
Weight Sensitivity of Temporal SNR Metrics in multi-echo fMRI
Purpose: In multi-echo fMRI (ME-fMRI), various weighting schemes have been proposed for the combination of the data across echoes. Here we introduce a framework that facilitates a deeper understanding of the weight dependence of temporal SNR measures in ME-fMRI. Theory and Methods: We examine two metrics that have been used to characterize ME-fMRI performance: temporal SNR (tSNR) and multi-echo temporal (metSNR). Both metrics can be described using the generalized Rayleigh quotient (GRQ) and are predicted to be relatively insensitive to the weights when there is a high degree of similarity between a metric-specific matrix in the GRQ numerator and a metricindependent covariance matrix in the GRQ denominator. The application of the GRQ framework to experimental data is demonstrated using a resting-state fMRI dataset acquired with a multi-echo multi-band EPI sequence. Results: In the example dataset, similarities between the covariance matrix and the metSNR and tSNR numerator matrices are highest in grey matter (GM) and cerebrospinal fluid (CSF) voxels, respectively. For representative GM and CSF voxels that exhibit high matrix similarity values, the metSNR and tSNR values, respectively, are both within 4% of their optimal values across a range of weighting schemes. However, there is a fundamental tradeoff, with a high degree of weight sensitivity in the tSNR and metSNR metrics for the representative GM and CSF voxels, respectively. Geometric insight into the observed weight dependencies is provided through a graphical interpretation of the GRQ. Conclusion: A GRQ framework can provide insight into the factors that determine the weight sensitivity of tSNR and metSNR measures in ME-fMRI.
Open AccessWeight Sensitivity of Temporal SNR Metrics in multi-echo fMRI
Purpose: In multi-echo fMRI (ME-fMRI), various weighting schemes have been proposed for the combination of the data across echoes. Here we introduce a framework that facilitates a deeper understanding of the weight dependence of temporal SNR measures in ME-fMRI. Theory and Methods: We examine two metrics that have been used to characterize ME-fMRI performance: temporal SNR (tSNR) and multi-echo temporal (metSNR). Both metrics can be described using the generalized Rayleigh quotient (GRQ) and are predicted to be relatively insensitive to the weights when there is a high degree of similarity between a metric-specific matrix in the GRQ numerator and a metricindependent covariance matrix in the GRQ denominator. The application of the GRQ framework to experimental data is demonstrated using a resting-state fMRI dataset acquired with a multi-echo multi-band EPI sequence. Results: In the example dataset, similarities between the covariance matrix and the metSNR and tSNR numerator matrices are highest in grey matter (GM) and cerebrospinal fluid (CSF) voxels, respectively. For representative GM and CSF voxels that exhibit high matrix similarity values, the metSNR and tSNR values, respectively, are both within 4% of their optimal values across a range of weighting schemes. However, there is a fundamental tradeoff, with a high degree of weight sensitivity in the tSNR and metSNR metrics for the representative GM and CSF voxels, respectively. Geometric insight into the observed weight dependencies is provided through a graphical interpretation of the GRQ. Conclusion: A GRQ framework can provide insight into the factors that determine the weight sensitivity of tSNR and metSNR measures in ME-fMRI.