Journal: PLoS ONE

Article Title: Reliability of multi-site UK Biobank MRI brain phenotypes for the assessment of neuropsychiatric complications of SARS-CoV-2 infection: The COVID-CNS travelling heads study

doi: 10.1371/journal.pone.0273704

Figure Lengend Snippet: In the top two panels, each column represents results for a different class of IDP, from left to right: regional and tissue volumes, cortical area, cortical thickness, regional and tissue intensity, and cortical grey-white contrast. A) Distribution of log-transformed P -values from repeated measures ANOVA testing for a site effect on the mean value of individual IDPs in each class; the solid horizontal line represents the P -value equivalent to FDR = 5%. Green dots represent IDPs fitted to the ANOVA model including data from all four sites; orange dots represent P -values for each IDP fitted to the ANOVA including only data from the three Siemens sites (Cambridge, Oxford, Liverpool). There are more significant between-site differences in mean IDPs, across all 5 classes, when the GE data from KCL are included in the analysis B) Swarm plots showing distribution of intra-class correlation coefficients (ICCs) for the same IDPs, estimated for each pair of all 4 sites (green points), for each pair of the three Siemens sites (orange points), and for comparable test-retest data drawn from the UKB cohort (blue points). Between-site reliability was generally high for all IDP classes compared to the UKB benchmark, whether or not GE data were included in the analysis. C) Each column represents finer-grained results for representative IDPs from each class of IDP: from left to right, left olfactory bulb volume, left precuneus area, left inferior temporal cortical thickness, left caudate intensity and left fusiform CNR. Top row , plots of each IDP for 8 subjects (coloured lines) scanned at each of 4 sites (x-axis labels); the grey violin plots indicate the distributions of the corresponding IDP in the UK Biobank reference datase, using matched random sampling of N = 8 participants. Box and whiskers represent inter-quartile range and 95% confidence intervals respectively. Bottom row , correlations between each pair of sites for each IDP: upper triangle, Pearson’s correlations; lower triangle, Spearman’s correlations.

Article Snippet: QSM-derived IDPs had generally better between-site reliability (all sites: ICC = 0.67, SD = 0.13; Siemens only sites: ICC = 0.76, SD = 0.14), comparable to good ctest-retest reliability in the UKB data (ICC = 0.66).

Techniques: Transformation Assay, Sampling

Journal: PLoS ONE

Article Title: Reliability of multi-site UK Biobank MRI brain phenotypes for the assessment of neuropsychiatric complications of SARS-CoV-2 infection: The COVID-CNS travelling heads study

doi: 10.1371/journal.pone.0273704

Figure Lengend Snippet: A) Representative T2 FLAIR images of the same subject scanned at each of 4 sites in the travelling heads study. B) left panel , peri-ventricular white matter hyperintensity volume for 8 subjects (coloured lines) scanned at each of 4 sites (x-axis labels); right panel , correlations between each pair of sites. C) left panel , deep white matter hyperintensity volume for 8 subjects (coloured lines) scanned at each of 4 sites (x-axis labels); right panel , correlations between each pair of sites. In both B) and C) , the upper triangle of the matrix shows Pearson’s correlations and the lower triangle shows Spearman’s correlations; and both IDPs were estimated using BIANCA.

Article Snippet: QSM-derived IDPs had generally better between-site reliability (all sites: ICC = 0.67, SD = 0.13; Siemens only sites: ICC = 0.76, SD = 0.14), comparable to good ctest-retest reliability in the UKB data (ICC = 0.66).

Techniques:

Journal: PLoS ONE

Article Title: Reliability of multi-site UK Biobank MRI brain phenotypes for the assessment of neuropsychiatric complications of SARS-CoV-2 infection: The COVID-CNS travelling heads study

doi: 10.1371/journal.pone.0273704

Figure Lengend Snippet: In the top two panels, the left column shows data for 14 IDPs derived from T2* data and the right column shows data for 14 IDPs derived from QSM data. A) Distribution of log-transformed P -values from repeated measures ANOVA testing for a site effect on the mean value of individual IDPs in each class; the solid horizontal line represents the P-value equivalent to FDR = 5%. Green dots represent IDPs fitted to the ANOVA model including data from all four sites; orange dots represent P -values for each IDP fitted to the ANOVA including only data from the three Siemens sites (Cambridge, Oxford, Liverpool). There were more significant between-site differences in mean IDPs when the GE data from KCL were included in the analysis B) Swarm plots showing distribution of intra-class correlation coefficients (ICCs) for the same IDPs, estimated for each pair of all 4 sites (green points), and for each pair of the three Siemens sites (orange points). C) Each column represents finer-grained results for representative IDPs from each class of IDP: from left to right, T2* right pallidum, QSM right pallidum. Top row , plots of each IDP for 8 subjects (coloured lines) scanned at each of 4 sites (x-axis labels). Bottom row , correlations between each pair of sites for each IDP: upper triangle, Pearson’s correlations; lower triangle, Spearman’s correlations.

Article Snippet: QSM-derived IDPs had generally better between-site reliability (all sites: ICC = 0.67, SD = 0.13; Siemens only sites: ICC = 0.76, SD = 0.14), comparable to good ctest-retest reliability in the UKB data (ICC = 0.66).

Techniques: Derivative Assay, Transformation Assay

Journal: PLoS ONE

Article Title: Reliability of multi-site UK Biobank MRI brain phenotypes for the assessment of neuropsychiatric complications of SARS-CoV-2 infection: The COVID-CNS travelling heads study

doi: 10.1371/journal.pone.0273704

Figure Lengend Snippet: In the top two panels, each column represents results for a different class of IDP, from left to right: white matter (WM) tract FA, WM tract MO, WM tract diffusivity, WM tract ICVF, WM tract OD and WM tract ISOVF. A) Distribution of log-transformed P -values from repeated measures ANOVA testing for a site effect on the mean value of individual IDPs in each class; the solid horizontal line represents the P -value equivalent to FDR = 5%. Green dots represent IDPs fitted to the ANOVA model including data from all four sites; orange dots represent P -values for each IDP fitted to the ANOVA including only data from the three Siemens sites (Cambridge, Oxford, Liverpool). There were more significant between-site differences in mean IDPs, across all 5 classes, when the GE data from KCL were included in the analysis. B) Swarm plots showing distribution of intra-class correlation coefficients (ICCs) for the same IDPs, estimated for each pair of all 4 sites (green points), for each pair of the three Siemens sites (orange points) and for comparable test-retest data drawn from the UKB cohort (blue points). Between-site reliability was generally high for all IDP classes compared to the UKB benchmark when only Siemens sites were included in the analysis. C) Each column represents finer-grained results for representative IDPs from each class of IDP: from left to right, FA right anterior thalamic radiation. MO left corona radiata, L3 left cingulate gyrus, ICVF left cingulate gyrus, OD superior cerebellar peduncle, and ISOVF superior longitudinal fasciculus. Top row , plots of each IDP for 8 subjects (coloured lines) scanned at each of 4 sites (x-axis labels); the grey violin plots indicate the distributions of the corresponding IDP in the UK Biobank reference dataset, using matched random sampling of N = 8 participants. Box and whiskers represent inter-quartile range and 95% confidence intervals respectively. Bottom row , correlations between each pair of sites for each IDP: upper triangle, Pearson’s correlations; lower triangle, Spearman’s correlations.

Article Snippet: QSM-derived IDPs had generally better between-site reliability (all sites: ICC = 0.67, SD = 0.13; Siemens only sites: ICC = 0.76, SD = 0.14), comparable to good ctest-retest reliability in the UKB data (ICC = 0.66).

Techniques: Transformation Assay, Sampling

Journal: PLoS ONE

Article Title: Reliability of multi-site UK Biobank MRI brain phenotypes for the assessment of neuropsychiatric complications of SARS-CoV-2 infection: The COVID-CNS travelling heads study

doi: 10.1371/journal.pone.0273704

Figure Lengend Snippet: The two columns show data on fMRI node amplitude and fMRI connectivity IDPs. Both represent IDPs derived from 25- and 100-node ICA-based parcellations. The fMRI connectivity IDPs represent 6 modes of variation across the functional connectivity network matrices derived from both parcellations. A ) Distribution of log-transformed P -values from repeated measures ANOVA testing for a site effect on the mean value of individual IDPs in each class; the solid horizontal line represents the P -value equivalent to FDR = 5%. Green dots represent IDPs fitted to the ANOVA model including data from all four sites; orange dots represent P -values for each IDP fitted to the ANOVA including only data from the three Siemens sites (Cambridge, Oxford, Liverpool). B) Swarm plots showing distribution of intra-class correlation coefficients (ICCs) for the same IDPs, estimated for each pair of all 4 sites (green points), for each pair of the three Siemens sites (orange points) and for comparable test-retest data drawn from the UKB cohort (blue points). Between-site reliability was generally high for all IDP classes compared to the UKB benchmark, whether or not GE data were included in the analysis. C) Each column represents finer-grained results for representative IDPs from each class of IDP: from left to right, fMRI node 4/25 (medial visual RSN) and summary connectivity mode #3 . Top row , plots of each IDP for 8 subjects (coloured lines) scanned at each of 4 sites (x-axis labels); the grey violin plot indicates the distribution of the corresponding IDP in the UK Biobank reference dataset. Bottom row , correlations between each pair of sites for each IDP: upper triangle, Pearson’s correlations; lower triangle, Spearman’s correlations.

Article Snippet: QSM-derived IDPs had generally better between-site reliability (all sites: ICC = 0.67, SD = 0.13; Siemens only sites: ICC = 0.76, SD = 0.14), comparable to good ctest-retest reliability in the UKB data (ICC = 0.66).

Techniques: Derivative Assay, Functional Assay, Transformation Assay