Population heterogeneity in clinical cohorts affects the predictive accuracy of brain imaging. Benkarim, O., Paquola, C., Park, B., Kebets, V., Hong, S., Wael, R. V. d., Zhang, S., Yeo, B. T. T., Eickenberg, M., Ge, T., Poline, J., Bernhardt, B. C., & Bzdok, D. PLOS Biology, 20(4):e3001627, April, 2022. Publisher: Public Library of Science![Population heterogeneity in clinical cohorts affects the predictive accuracy of brain imaging [link]](https://bibbase.org/img/filetypes/link.svg "link")
Paper doi abstract bibtex Brain imaging research enjoys increasing adoption of supervised machine learning for single-participant disease classification. Yet, the success of these algorithms likely depends on population diversity, including demographic differences and other factors that may be outside of primary scientific interest. Here, we capitalize on propensity scores as a composite confound index to quantify diversity due to major sources of population variation. We delineate the impact of population heterogeneity on the predictive accuracy and pattern stability in 2 separate clinical cohorts: the Autism Brain Imaging Data Exchange (ABIDE, n = 297) and the Healthy Brain Network (HBN, n = 551). Across various analysis scenarios, our results uncover the extent to which cross-validated prediction performances are interlocked with diversity. The instability of extracted brain patterns attributable to diversity is located preferentially in regions part of the default mode network. Collectively, our findings highlight the limitations of prevailing deconfounding practices in mitigating the full consequences of population diversity.
@article{benkarim_population_2022,
title = {Population heterogeneity in clinical cohorts affects the predictive accuracy of brain imaging},
volume = {20},
issn = {1545-7885},
url = {https://journals.plos.org/plosbiology/article?id=10.1371/journal.pbio.3001627},
doi = {10.1371/journal.pbio.3001627},
abstract = {Brain imaging research enjoys increasing adoption of supervised machine learning for single-participant disease classification. Yet, the success of these algorithms likely depends on population diversity, including demographic differences and other factors that may be outside of primary scientific interest. Here, we capitalize on propensity scores as a composite confound index to quantify diversity due to major sources of population variation. We delineate the impact of population heterogeneity on the predictive accuracy and pattern stability in 2 separate clinical cohorts: the Autism Brain Imaging Data Exchange (ABIDE, n = 297) and the Healthy Brain Network (HBN, n = 551). Across various analysis scenarios, our results uncover the extent to which cross-validated prediction performances are interlocked with diversity. The instability of extracted brain patterns attributable to diversity is located preferentially in regions part of the default mode network. Collectively, our findings highlight the limitations of prevailing deconfounding practices in mitigating the full consequences of population diversity.},
language = {en},
number = {4},
urldate = {2022-05-04},
journal = {PLOS Biology},
author = {Benkarim, Oualid and Paquola, Casey and Park, Bo-yong and Kebets, Valeria and Hong, Seok-Jun and Wael, Reinder Vos de and Zhang, Shaoshi and Yeo, B. T. Thomas and Eickenberg, Michael and Ge, Tian and Poline, Jean-Baptiste and Bernhardt, Boris C. and Bzdok, Danilo},
month = apr,
year = {2022},
note = {Publisher: Public Library of Science},
keywords = {ADHD, Autism, Autism spectrum disorder, Forecasting, Machine learning, Neural networks, Neuroimaging, Species diversity},
pages = {e3001627},
}
Downloads: 0
{"_id":"5Qaquz8fjYuayzNcX","bibbaseid":"benkarim-paquola-park-kebets-hong-wael-zhang-yeo-etal-populationheterogeneityinclinicalcohortsaffectsthepredictiveaccuracyofbrainimaging-2022","author_short":["Benkarim, O.","Paquola, C.","Park, B.","Kebets, V.","Hong, S.","Wael, R. V. d.","Zhang, S.","Yeo, B. T. T.","Eickenberg, M.","Ge, T.","Poline, J.","Bernhardt, B. C.","Bzdok, D."],"bibdata":{"bibtype":"article","type":"article","title":"Population heterogeneity in clinical cohorts affects the predictive accuracy of brain imaging","volume":"20","issn":"1545-7885","url":"https://journals.plos.org/plosbiology/article?id=10.1371/journal.pbio.3001627","doi":"10.1371/journal.pbio.3001627","abstract":"Brain imaging research enjoys increasing adoption of supervised machine learning for single-participant disease classification. Yet, the success of these algorithms likely depends on population diversity, including demographic differences and other factors that may be outside of primary scientific interest. Here, we capitalize on propensity scores as a composite confound index to quantify diversity due to major sources of population variation. We delineate the impact of population heterogeneity on the predictive accuracy and pattern stability in 2 separate clinical cohorts: the Autism Brain Imaging Data Exchange (ABIDE, n = 297) and the Healthy Brain Network (HBN, n = 551). Across various analysis scenarios, our results uncover the extent to which cross-validated prediction performances are interlocked with diversity. The instability of extracted brain patterns attributable to diversity is located preferentially in regions part of the default mode network. Collectively, our findings highlight the limitations of prevailing deconfounding practices in mitigating the full consequences of population diversity.","language":"en","number":"4","urldate":"2022-05-04","journal":"PLOS Biology","author":[{"propositions":[],"lastnames":["Benkarim"],"firstnames":["Oualid"],"suffixes":[]},{"propositions":[],"lastnames":["Paquola"],"firstnames":["Casey"],"suffixes":[]},{"propositions":[],"lastnames":["Park"],"firstnames":["Bo-yong"],"suffixes":[]},{"propositions":[],"lastnames":["Kebets"],"firstnames":["Valeria"],"suffixes":[]},{"propositions":[],"lastnames":["Hong"],"firstnames":["Seok-Jun"],"suffixes":[]},{"propositions":[],"lastnames":["Wael"],"firstnames":["Reinder","Vos","de"],"suffixes":[]},{"propositions":[],"lastnames":["Zhang"],"firstnames":["Shaoshi"],"suffixes":[]},{"propositions":[],"lastnames":["Yeo"],"firstnames":["B.","T.","Thomas"],"suffixes":[]},{"propositions":[],"lastnames":["Eickenberg"],"firstnames":["Michael"],"suffixes":[]},{"propositions":[],"lastnames":["Ge"],"firstnames":["Tian"],"suffixes":[]},{"propositions":[],"lastnames":["Poline"],"firstnames":["Jean-Baptiste"],"suffixes":[]},{"propositions":[],"lastnames":["Bernhardt"],"firstnames":["Boris","C."],"suffixes":[]},{"propositions":[],"lastnames":["Bzdok"],"firstnames":["Danilo"],"suffixes":[]}],"month":"April","year":"2022","note":"Publisher: Public Library of Science","keywords":"ADHD, Autism, Autism spectrum disorder, Forecasting, Machine learning, Neural networks, Neuroimaging, Species diversity","pages":"e3001627","bibtex":"@article{benkarim_population_2022,\n\ttitle = {Population heterogeneity in clinical cohorts affects the predictive accuracy of brain imaging},\n\tvolume = {20},\n\tissn = {1545-7885},\n\turl = {https://journals.plos.org/plosbiology/article?id=10.1371/journal.pbio.3001627},\n\tdoi = {10.1371/journal.pbio.3001627},\n\tabstract = {Brain imaging research enjoys increasing adoption of supervised machine learning for single-participant disease classification. Yet, the success of these algorithms likely depends on population diversity, including demographic differences and other factors that may be outside of primary scientific interest. Here, we capitalize on propensity scores as a composite confound index to quantify diversity due to major sources of population variation. We delineate the impact of population heterogeneity on the predictive accuracy and pattern stability in 2 separate clinical cohorts: the Autism Brain Imaging Data Exchange (ABIDE, n = 297) and the Healthy Brain Network (HBN, n = 551). Across various analysis scenarios, our results uncover the extent to which cross-validated prediction performances are interlocked with diversity. The instability of extracted brain patterns attributable to diversity is located preferentially in regions part of the default mode network. Collectively, our findings highlight the limitations of prevailing deconfounding practices in mitigating the full consequences of population diversity.},\n\tlanguage = {en},\n\tnumber = {4},\n\turldate = {2022-05-04},\n\tjournal = {PLOS Biology},\n\tauthor = {Benkarim, Oualid and Paquola, Casey and Park, Bo-yong and Kebets, Valeria and Hong, Seok-Jun and Wael, Reinder Vos de and Zhang, Shaoshi and Yeo, B. T. Thomas and Eickenberg, Michael and Ge, Tian and Poline, Jean-Baptiste and Bernhardt, Boris C. and Bzdok, Danilo},\n\tmonth = apr,\n\tyear = {2022},\n\tnote = {Publisher: Public Library of Science},\n\tkeywords = {ADHD, Autism, Autism spectrum disorder, Forecasting, Machine learning, Neural networks, Neuroimaging, Species diversity},\n\tpages = {e3001627},\n}\n\n","author_short":["Benkarim, O.","Paquola, C.","Park, B.","Kebets, V.","Hong, S.","Wael, R. V. d.","Zhang, S.","Yeo, B. T. T.","Eickenberg, M.","Ge, T.","Poline, J.","Bernhardt, B. C.","Bzdok, D."],"key":"benkarim_population_2022","id":"benkarim_population_2022","bibbaseid":"benkarim-paquola-park-kebets-hong-wael-zhang-yeo-etal-populationheterogeneityinclinicalcohortsaffectsthepredictiveaccuracyofbrainimaging-2022","role":"author","urls":{"Paper":"https://journals.plos.org/plosbiology/article?id=10.1371/journal.pbio.3001627"},"keyword":["ADHD","Autism","Autism spectrum disorder","Forecasting","Machine learning","Neural networks","Neuroimaging","Species diversity"],"metadata":{"authorlinks":{}}},"bibtype":"article","biburl":"https://api.zotero.org/groups/4553800/items?key=BfP7bN7FF9dJwtyiLBORewdg&format=bibtex&limit=100","dataSources":["5EcJ4DjTLGyJLTga3"],"keywords":["adhd","autism","autism spectrum disorder","forecasting","machine learning","neural networks","neuroimaging","species diversity"],"search_terms":["population","heterogeneity","clinical","cohorts","affects","predictive","accuracy","brain","imaging","benkarim","paquola","park","kebets","hong","wael","zhang","yeo","eickenberg","ge","poline","bernhardt","bzdok"],"title":"Population heterogeneity in clinical cohorts affects the predictive accuracy of brain imaging","year":2022}