Image Quality Variation and Cortical Volume Estimation in a Multisite Study of Low Back Pain
Poster No:
2237
Submission Type:
Abstract Submission
Authors:
Jennifer Cummings1, Jagan Jimmy2, Richard Harris3, Eric Ichesco2, Chelsea Kaplan2, Salvatore Torrisi1, An Vu1, Hsiao-Ying Wey4, Chi-Hyeon Yoo4, Scott Peltier2
Institutions:
1University of California, San Francisco, San Francisco, CA, 2University of Michigan, Ann Arbor, MI, 3University of California at Irvine, Irvine, CA, 4Massachusetts General Hospital, Boston, MA
First Author:
Co-Author(s):
Introduction:
Chronic low back pain (cLBP) is one of the leading causes of disability in the world and one of the top non-cancer reasons for opioid prescription in the United States (Ferreira 2023; Ringwalt 2014). The Back Pain Consortium (BACPAC) Research Program is an interdisciplinary effort to better understand the biopsychosocial factors contributing to cLBP (Mauck 2023). Within the Consortium, the Brain Imaging Working Group conducts neuroimaging studies aimed at elucidating brain structural and functional biomarkers associated with pain perception and treatment efficacy.
To collect a large, diverse participant sample, a protocol was developed to standardize image acquisition parameters across the different hardware and software configurations at participating sites. This work presents preliminary results comparing image quality control (QC) metrics and morphometric measures derived from structural MR images collected on 10 MR scanners within the BACPAC network.
To collect a large, diverse participant sample, a protocol was developed to standardize image acquisition parameters across the different hardware and software configurations at participating sites. This work presents preliminary results comparing image quality control (QC) metrics and morphometric measures derived from structural MR images collected on 10 MR scanners within the BACPAC network.
Methods:
Overview of Participants
Participants were recruited as part of each institution's BACPAC study. Common inclusion criteria include the presence of low back pain. Here we report results from the first 8-16 subjects per site. In addition, a traveling subject (healthy volunteer, male, age 28-29, BMI = 27.9) was scanned on at least one scanner at each site between July 2021 and September 2022.
Image Acquisition
Neuroimaging was performed on a total of 10 different scanners across 6 sites. The scan protocol outlined sequences for three MRI techniques: T1-weighted and T2-weighted structural imaging and functional MRI using single shot echo planar imaging. Here we report results using T1-weighted images only (MP-RAGE, TR/TE 2500/2.88 (Siemens), 2500/2 (GE); voxel size 1.0 mm isotropic; matrix 256x256; flip angle 8 deg). All sites used 3T MR Siemens or GE scanners.
Quality Control
Image quality metrics were extracted using MRIQC v23.1.0 (Esteban, 2017). Here we report 4 summary metrics: the signal-to-noise ratio calculated within the tissue mask (SNR_Total); the contrast-to-noise ratio representing the separation of gray and white matter signals (CNR); the average full width at half maximum, a measure of image smoothness (FWHM_AVG); and a ratio of the SNR to FWHM (SNR/FWHM). All metrics and formulas are described in further detail in MRIQC's documentation. One-way ANOVA was performed to test for site differences in each of the QC metrics. Tests with a p-value less than 0.05 are reported.
Brain Volume Analysis
Brain volume measurements were derived from traveling subject data using Freesurfer v1.201 (Dale 1999). Here we report total brain volume and insula volume, an ROI involved in pain processing (Labrakakis, 2023).
Participants were recruited as part of each institution's BACPAC study. Common inclusion criteria include the presence of low back pain. Here we report results from the first 8-16 subjects per site. In addition, a traveling subject (healthy volunteer, male, age 28-29, BMI = 27.9) was scanned on at least one scanner at each site between July 2021 and September 2022.
Image Acquisition
Neuroimaging was performed on a total of 10 different scanners across 6 sites. The scan protocol outlined sequences for three MRI techniques: T1-weighted and T2-weighted structural imaging and functional MRI using single shot echo planar imaging. Here we report results using T1-weighted images only (MP-RAGE, TR/TE 2500/2.88 (Siemens), 2500/2 (GE); voxel size 1.0 mm isotropic; matrix 256x256; flip angle 8 deg). All sites used 3T MR Siemens or GE scanners.
Quality Control
Image quality metrics were extracted using MRIQC v23.1.0 (Esteban, 2017). Here we report 4 summary metrics: the signal-to-noise ratio calculated within the tissue mask (SNR_Total); the contrast-to-noise ratio representing the separation of gray and white matter signals (CNR); the average full width at half maximum, a measure of image smoothness (FWHM_AVG); and a ratio of the SNR to FWHM (SNR/FWHM). All metrics and formulas are described in further detail in MRIQC's documentation. One-way ANOVA was performed to test for site differences in each of the QC metrics. Tests with a p-value less than 0.05 are reported.
Brain Volume Analysis
Brain volume measurements were derived from traveling subject data using Freesurfer v1.201 (Dale 1999). Here we report total brain volume and insula volume, an ROI involved in pain processing (Labrakakis, 2023).
Results:
Image quality metrics for all subject data are shown in Figure 1. There is a statistically significant difference between the sites for all four QC metrics, as follows:
SNR_Total: F= 13.17, p-value= 6.13e-10, with an overall large effect (ω2) = 0.36.
CNR: F= 11.22, p-value= 1.15e-8, with an overall large effect (ω2)= 0.32.
FWHM_AVG: F= 13.00, p-value= 7.92e-10, with an overall large effect (ω2) = 0.35.
SNR/FWHM: F= 7.05, p-value= 0.000010, with an overall large effect (ω2) = 0.22.
Results of the brain volume analysis are shown in Figure 2. Whole brain volume estimations for the traveling subject = 1289874.43 ± 13209.57 mm3; Insula volume estimations = 7094.86 ± 271.21 mm3.
SNR_Total: F= 13.17, p-value= 6.13e-10, with an overall large effect (ω2) = 0.36.
CNR: F= 11.22, p-value= 1.15e-8, with an overall large effect (ω2)= 0.32.
FWHM_AVG: F= 13.00, p-value= 7.92e-10, with an overall large effect (ω2) = 0.35.
SNR/FWHM: F= 7.05, p-value= 0.000010, with an overall large effect (ω2) = 0.22.
Results of the brain volume analysis are shown in Figure 2. Whole brain volume estimations for the traveling subject = 1289874.43 ± 13209.57 mm3; Insula volume estimations = 7094.86 ± 271.21 mm3.
·Each subject is represented by a black point and where available, corresponding traveling subject data (ts) is represented in yellow. Boxes are colored according to the scanner manufacturer.
·Brain volume measurements derived from traveling subject data. Each point represents the same subject’s acquisition at a different scanner and is colored according to the legend on the right.
Conclusions:
We show that group level image quality measures vary significantly between sites despite harmonization of the imaging protocol. We also show variation of several hundred mm3 in brain volume estimation on the traveling subject when scanned at different sites. Future directions will include a deeper investigation into the causes of these variations, such as image artifacts, coil or software differences, as well as the impacts and considerations when pooling the data for analysis.
Modeling and Analysis Methods:
Image Registration and Computational Anatomy 2
Other Methods
Neuroinformatics and Data Sharing:
Databasing and Data Sharing 1
Novel Imaging Acquisition Methods:
Anatomical MRI
Perception, Attention and Motor Behavior:
Perception: Pain and Visceral
Keywords:
Acquisition
Design and Analysis
Experimental Design
MRI
Pain
STRUCTURAL MRI
Other - Quality Control; Data Harmonization
1|2Indicates the priority used for review
Provide references using author date format
Dale, A. M. (1999), 'Cortical surface-based analysis: I. Segmentation and surface reconstruction', Neuroimage, 9(2), 179-194.
Esteban, O. (2017), 'MRIQC: Advancing the automatic prediction of image quality in MRI from unseen sites', PloS one, 12(9), e0184661.
Ferreira, M. L. (2023), 'Global, regional, and national burden of low back pain, 1990–2020, its attributable risk factors, and projections to 2050: a systematic analysis of the Global Burden of Disease Study 2021', The Lancet Rheumatology, 5(6), e316-e329.
Labrakakis, C. (2023), 'The Role of the insular Cortex in Pain', International Journal of Molecular Sciences, 24(6), 5736.
Mauck, M. C. (2023), 'The Back Pain Consortium (BACPAC) Research Program: Structure, Research Priorities, and Methods', Pain Medicine, pnac202.
Ringwalt, C. (2014), 'Differential prescribing of opioid analgesics according to physician specialty for Medicaid patients with chronic noncancer pain diagnoses', Pain Research and Management, 19, 179-185.