ceitec_PPT_podklad_uvod logo+napis_en OPVaVpI_loga-eu_pos_H_EN Meta-analysis of neuroimaging data Martin Jáni Selected Topics in Contemporary Neuroscience CEITEC_logo_neg_RGB https://www.frontiersin.org/files/Articles/89718/fnhum-08-00462-HTML/image_m/fnhum-08-00462-g001.jp g [Stelzer et al., 2014] CEITEC_logo_neg_RGB Súvisiaci obrázok CEITEC_logo_neg_RGB why and what? •need for comprehensive summary * meta-analysis = quantitative review * 1 study represents 1 subject * units of measurement = summary statistics (effect sizes) * standard IMRaD structure CEITEC_logo_neg_RGB SouvisejÃcà obrázek SouvisejÃcà obrázek SouvisejÃcà obrázek SouvisejÃcà obrázek •t-statistic •p value CEITEC_logo_neg_RGB SouvisejÃcà obrázek SouvisejÃcà obrázek SouvisejÃcà obrázek SouvisejÃcà obrázek SouvisejÃcà obrázek SouvisejÃcà obrázek SouvisejÃcà obrázek SouvisejÃcà obrázek SouvisejÃcà obrázek CEITEC_logo_neg_RGB The use of Effect sizes * difference in means / pooled standard deviation •Cohen's d •Glass' Δ •Hedges’ g * comparable with other studies * unit of measure is lost * less straightforward http://www.statisticshowto.com/wp-content/uploads/2016/10/hedges-g-pooled-weighted.png CEITEC_logo_neg_RGB Publication bias http://media.springernature.com/full/springer-static/image/art%3A10.1186%2F2045-5380-2-6/MediaObjec ts/13587_2011_Article_18_Fig2_HTML.jpg [Radua and Mataix-Cols, 2012] •Forest plot •Funnel plot CEITEC_logo_neg_RGB •Region of interest-based meta-analyses Výsledok vyhľadávania obrázkov pre dopyt region of interest Voxel-based meta-analyses Výsledok vyhľadávania obrázkov pre dopyt voxel wise CEITEC_logo_neg_RGB Region of interest-based meta-analyses * set of different meta-analyses for every ROI * selective = some regions more studied than others * a priori hypotheses influence selection of ROI * strong publication bias Výsledok vyhľadávania obrázkov pre dopyt region of interest CEITEC_logo_neg_RGB Label-based reviews •peak of a cluster plotted as a dot •number of dots count in each region * increase * decrease * borders of conventional regions • •drawbacks: no weighting, loss of information, selective * http://media.springernature.com/full/springer-static/image/art%3A10.1186%2F2045-5380-2-6/MediaObjec ts/13587_2011_Article_18_Fig3_HTML.jpg [Radua and Mataix-Cols, 2012] CEITEC_logo_neg_RGB Voxel-based meta-analyses * Image-based meta-analyses * Coordinate-based meta-analyses * Mixed image- and coordinate-based meta-analyses * CEITEC_logo_neg_RGB Voxel-based meta-analyses * Image-based meta-analyses * Coordinate-based meta-analyses * Mixed image- and coordinate-based meta-analyses CEITEC_logo_neg_RGB Image-based meta-analyses * use of parametric maps * meta-analysis for each voxel * multiple-comparisons problem * hard to find (contacting authors) Výsledok vyhľadávania obrázkov pre dopyt voxel wise CEITEC_logo_neg_RGB Voxel-based meta-analyses * Image-based meta-analyses * Coordinate-based meta-analyses * Mixed image- and coordinate-based meta-analyses CEITEC_logo_neg_RGB Coordinate-based meta-analyses •Kanske et al. 2015 CEITEC_logo_neg_RGB Kernel density analysis (KDA) http://media.springernature.com/full/springer-static/image/art%3A10.1186%2F2045-5380-2-6/MediaObjec ts/13587_2011_Article_18_Fig3_HTML.jpg •peak as a sphere •number of spheres surrounding each voxel are counted •increase •decrease • • • [Radua and Mataix-Cols, 2012] Peak coordinates mapping in some voxel-based meta-analytic methods CEITEC_logo_neg_RGB Multilevel kernel density (MKDA) Peak coordinates mapping in some voxel-based meta-analytic methods http://media.springernature.com/full/springer-static/image/art%3A10.1186%2F2045-5380-2-6/MediaObjec ts/13587_2011_Article_18_Fig3_HTML.jpg •similar to KDA •voxel close to two spheres from one study counts as one •avoids false high values at intersections •weighted by sample size •robustness analysis [Radua and Mataix-Cols, 2012] CEITEC_logo_neg_RGB Activation likelihood estimation (ALE) http://media.springernature.com/full/springer-static/image/art%3A10.1186%2F2045-5380-2-6/MediaObjec ts/13587_2011_Article_18_Fig3_HTML.jpg •peak as a smoothed sphere (Gaussian Kernel at FWHM) •higher value for voxels closer to the center of the sphere (peak) •increase •decrease • • • [Radua and Mataix-Cols, 2012] Peak coordinates mapping in some voxel-based meta-analytic methods CEITEC_logo_neg_RGB Signed differential mapping (SDM) http://media.springernature.com/full/springer-static/image/art%3A10.1186%2F2045-5380-2-6/MediaObjec ts/13587_2011_Article_18_Fig3_HTML.jpg •smoothed spheres like in ALE •weighted by sample size, robustness analysis like MKDA • •combines positive and negative values •adds heterogeneity analysis [Radua and Mataix-Cols, 2012] Peak coordinates mapping in some voxel-based meta-analytic methods CEITEC_logo_neg_RGB Peak coordinates mapping in some voxel-based meta-analytic methods Effect size-Signed differential mapping (ES-SDM) http://media.springernature.com/full/springer-static/image/art%3A10.1186%2F2045-5380-2-6/MediaObjec ts/13587_2011_Article_18_Fig3_HTML.jpg •similar to SDM, but values are effect sizes •weighted by variance (right) • [Radua and Mataix-Cols, 2012] random effects model combination of peaks and statistical parametric maps CEITEC_logo_neg_RGB Voxel-based meta-analyses Peak coordinates mapping in some voxel-based meta-analytic methods [Radua et al., 2012] CEITEC_logo_neg_RGB Voxel-based meta-analyses * Image-based meta-analyses * Coordinate-based meta-analyses * Mixed image- and coordinate-based meta-analyses CEITEC_logo_neg_RGB Mixed image- and coordinate-based meta-analyses Validation of ES-SDM [Radua et al., 2012] CEITEC_logo_neg_RGB Summary * Meta-analysis is a quantitative systematic review * Pretty accessible and valuable research to do * Different approaches and methods * ROI based and Voxel based CEITEC_logo_neg_RGB Example: SDM •www.sdmproject.com CEITEC_logo_neg_RGB Example: SDM 1.research question 2.data collection 3.formatting 4.preprocessing 5.model estimation 6.results •www.sdmproject.com CEITEC_logo_neg_RGB Example: SDM 1.research question 2.data collection 3.formatting 4.preprocessing 5.model estimation 6.results * consistent task/process * choose contrasts CEITEC_logo_neg_RGB Example: SDM 1.research question 2.data collection 3.formatting 4.preprocessing 5.model estimation 6.results * similar to systematic review * search in databases * selection of papers * 1 study as 1 subject * contact authors for missing data CEITEC_logo_neg_RGB Example: SDM 1.research question 2.data collection 3.formatting 4.preprocessing 5.model estimation 6.results CEITEC_logo_neg_RGB Example: SDM 1.research question 2.data collection 3.formatting 4.preprocessing 5.model estimation 6.results CEITEC_logo_neg_RGB Example: SDM 1.research question 2.data collection 3.formatting 4.preprocessing 5.model estimation 6.results CEITEC_logo_neg_RGB Example: SDM 1.research question 2.data collection 3.formatting 4.preprocessing 5.model estimation 6.results * convert coordinate peaks to estimate parametric maps * optionally include original parametric maps * * creates a)parametric maps of effect sizes b)heterogeneity maps • CEITEC_logo_neg_RGB Example: SDM 1.research question 2.data collection 3.formatting 4.preprocessing 5.model estimation 6.results * modality: VBM, fMRI, PET, DTI * template: GM, WM, FA, CSF CEITEC_logo_neg_RGB Example: SDM 1.research question 2.data collection 3.formatting 4.preprocessing 5.model estimation 6.results * estimate mean * linear model a)compare groups b)meta-regression * multimodal meta-analysis * CEITEC_logo_neg_RGB Example: SDM 1.research question 2.data collection 3.formatting 4.preprocessing 5.model estimation 6.results * threshold results * extract peaks (seed) * funnel plot CEITEC_logo_neg_RGB Example: SDM 1.research question 2.data collection 3.formatting 4.preprocessing 5.model estimation 6.results * threshold results * create mask * extract peaks * funnel plot CEITEC_logo_neg_RGB Example: SDM 1.research question 2.data collection 3.formatting 4.preprocessing 5.model estimation 6.results * threshold results * create mask * extract peaks * funnel plot CEITEC_logo_neg_RGB Databases * sets of original data (eg raw scanned images) • BRAINNet (http://www.brainnet.net) • fMRI Data Center (http://www.fmridc.org) • OpenfMRI (http://www.openfmri.org) * summary statistics from the studies included in one meta-analysis (mean and SD of ROI volumes) • Bipolar Disorder Neuroimaging Database (http://www.bipolardatabase.org) • Major Depressive Disorder Neuroimaging Database (http://www.depressiondatabase.org) • Peak-coordinate databases from SDM meta-analyses (http://www.sdmproject.com/database) * sets of summary statistics of virtually all published studies • BrainMap (http://www.brainmap.org) • NeuroSynth (http://www.neurosynth.org) ceitec_PPT_podklad_uvod CEITEC_logo_pos_RGB OPVaVpI_loga-eu_pos_H_EN Thank you for your attention CEITEC_logo_neg_RGB literature •Kanske P, Schönfelder S, Forneck J, Wessa M (2015): Impaired regulation of emotion: neural correlates of reappraisal and distraction in bipolar disorder and unaffected relatives. Transl Psychiatry 5:e497. http://www.nature.com/doifinder/10.1038/tp.2014.137. •Radua J, Mataix-Cols D, Phillips ML, El-Hage W, Kronhaus DM, Cardoner N, Surguladze S (2012): A new meta-analytic method for neuroimaging studies that combines reported peak coordinates and statistical parametric maps. Eur Psychiatry 27:605–11. http://www.sciencedirect.com/science/article/pii/S0924933811000733. •Radua J, Mataix-Cols D (2009): Voxel-wise meta-analysis of grey matter changes in obsessive–compulsive disorder. Br J Psychiatry 195:393–402. http://bjp.rcpsych.org/content/195/5/393.abstract. •Radua J, Mataix-Cols D (2012): Meta-analytic methods for neuroimaging data explained. Biol Mood Anxiety Disord 2:6. http://biolmoodanxietydisord.biomedcentral.com/articles/10.1186/2045-5380-2-6. •Stelzer J, Lohmann G, Mueller K, Buschmann T, Turner R (2014): Deficient approaches to human neuroimaging. Front Hum Neurosci 8:1–16. http://journal.frontiersin.org/article/10.3389/fnhum.2014.00462/abstract.