[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"doc-detail-120652-en":3,"doc-seo-120652-105":30,"detail-sidebar-cat-0-en-105":91},{"code":4,"msg":5,"data":6},0,"success",{"doc_id":7,"user_id":8,"nickname":9,"user_avatar":10,"doc_module":4,"category_id":11,"category_name":12,"doc_title":13,"doc_description":14,"doc_content":15,"file_id":16,"file_url":17,"file_type":18,"file_size":19,"view_count":20,"is_deleted":4,"is_public":20,"is_downloadable":20,"audit_status":20,"page_count":21,"language":22,"language_code":23,"site_id":24,"html_lang":23,"table_of_contents":25,"faqs":26,"seo_title":27,"seo_description":14,"update_tm":28,"read_time":29},120652,1099513958607,"Jiven","https://ap-avatar.wpscdn.com/avatar/100002390cf8733938c?x-image-process=image/resize,m_fixed,w_180,h_180&k=1778829742770036399",8,"Research & Report","Efficacy of MRI data harmonization in the age of machine learning - A multicenter study across 36 datasets","Pooling publicly available MRI data across multiple sites enables larger, more diverse subject cohorts, boosts statistical power, and supports data reuse for machine learning. Multicenter harmonization is required to reduce confounding from non-biological variability, but applying it to the full dataset before model training can cause data leakage and inflate performance estimates. This work measures harmonization efficacy and proposes a harmonizer transformer to integrate ComBat within the preprocessing pipeline, preventing leakage while preserving biological signals.","Efficacy of MRI data harmonization in the age of machine learning. A multicenter study across 36 datasets  \nChiara Marzi 1, Marco Giannelli2, Andrea Barucci 1, Carlo Tessa3, Mario Mascalchi4, Stefano Diciotti5,6  \n1“Nello Carrara” Institute of Applied Physics (IFAC), National Research Council (CNR), 50019 Sesto Fiorentino (Firenze), Italy  \n2Unit of Medical Physics, Pisa University Hospital “Azienda Ospedaliero-Universitaria Pisana”, 56126 Pisa, Italy  \n3Radiology Unit Apuane e Lunigiana, Azienda USL Toscana Nord Ovest, 54100 Massa, Italy  \n4Department of Experimental and Clinical Biomedical Sciences “Mario Serio”, University of Florence, 50139 Florence Italy  \n5Department of Electrical, Electronic, and Information Engineering “Guglielmo Marconi” -DEI, University of Bologna, 47522 Cesena, Italy  \n6Alma Mater Research Institute for Human-Centered Artificial Intelligence, University of Bologna, 40121 Bologna, Italy  \nCorresponding Author: Prof. Stefano Diciotti, Department of Electrical, Electronic, and Information Engineering “Guglielmo Marconi”, University of Bologna, Via dell’Università 50, 47521, Cesena, Italy. E-mail: [stefano.diciotti@unibo.it](stefano.diciotti@unibo.it)  \nAbstract  \nPooling publicly-available MRI data from multiple sites allows to assemble extensive groups of subjects, increase statistical power, and promote data reuse with machine learning techniques. The harmonization of multicenter data is necessary to reduce the confounding effect associated with non-biological sources of variability in the data. However, when applied to the entire dataset before machine learning, the harmonization leads to data leakage, because information outside the training set may affect model building, and potentially falsely overestimate performance. We propose a 1) measurement of the efficacy of data harmonization; 2) harmonizer transformer, i.e., an implementation of the ComBat harmonization allowing its encapsulation among the preprocessing steps ofa machine learning pipeline, avoiding data leakage. We tested these tools using brain T 1-weighted MRI data from 1740 healthy subjects acquired at 36 sites. After harmonization, the site effect was removed or reduced, and we measured the data leakage effect in predicting individual age from MRI data, highlighting that introducing the harmonizer transformer into a machine learning pipeline allows for avoiding data leakage.  \n1. Background & Summary  \nIn recent years there has been an increasing trend toward data sharing in neuroimaging research communities, leading to a rising number of public neuroimaging databases and collaborative multicenter initiatives1–4. Indeed, pooling MRI data from multiple sites providesan opportunity to assemble more extensive and diverse groups of subjects2,3,5,6, increase statistical power3,7–10, and study rare disorders and subtle effects 11,12 . However, a major drawback of combining neuroimaging data across sites is the introduction of confounding effects due to non-biological variability in the data, typically related to image acquisition hardware and protocol. Indeed, properties of MRI such as scanner field strength, radiofrequency coil type, gradients coil characteristics, hardware, image reconstruction algorithm, and non-standardized acquisition protocol parameters can introduce unwanted technical variability, also reflected in MRI-derived features13–15.  \nThe harmonization of multicenter data, defined as applying mathematical and statistical concepts to reduce unwanted site variability while maintaining the biological content, is, therefore, necessary to ensure the success of cooperative analyses. Currently, among the harmonization methods available to the neuroimaging scientific community (e.g., functional normalization 16, RAVEL, global scaling, adjusted residuals harmonization17), ComBat is one of the most widely used7,12,18–27 . The ComBat model was first introduced in gene expression analysis as a batch-effect correction tool to remove unw","cbCaigwVaK9wXGCh","https://ap.wps.com/l/cbCaigwVaK9wXGCh","pdf",2726760,1,56,"English","en",105,"# Background & Summary\n## Data sharing in neuroimaging\n## Confounding from site-related variability\n## Harmonization methods and ComBat\n## Data leakage risk in preprocessing\n## Study contribution and tools","[{\"question\":\"Why is MRI data harmonization needed in multicenter machine learning studies?\",\"answer\":\"Because MRI acquisition hardware and protocol parameters introduce non-biological site variability that can confound analyses. Harmonization reduces site effects while aiming to maintain biological content.\"},{\"question\":\"What problem arises when harmonization is applied to the entire dataset before training?\",\"answer\":\"Harmonization can lead to data leakage, since information from outside the training set may influence model building. This can falsely overestimate predictive performance.\"},{\"question\":\"What is the proposed harmonizer transformer and how does it help?\",\"answer\":\"It is an implementation that encapsulates ComBat harmonization within the preprocessing steps of a machine learning pipeline. By fitting harmonization parameters only on training data and applying them to test data, it avoids data leakage.\"}]","Efficacy of MRI data harmonization in the age of machine learning - A multicenter study across 36 datasets | PDF",1785731175,141,{"code":4,"msg":31,"data":32},"ok",{"site_id":24,"language":23,"slug":33,"title":13,"keywords":34,"description":14,"schema_data":35,"social_meta":86,"head_meta":88,"extra_data":90,"updated_unix":28},"efficacy-of-mri-data-harmonization-in-the-age-of-machine-learning-a-multicenter-study-across-36-datasets","",{"@graph":36,"@context":85},[37,54,68],{"@type":38,"itemListElement":39},"BreadcrumbList",[40,44,48,51],{"item":41,"name":42,"@type":43,"position":20},"https://docshare.wps.com","Home","ListItem",{"item":45,"name":46,"@type":43,"position":47},"https://docshare.wps.com/document/","Document",2,{"item":49,"name":12,"@type":43,"position":50},"https://docshare.wps.com/document/research-report/",3,{"item":52,"name":13,"@type":43,"position":53},"https://docshare.wps.com/document/efficacy-of-mri-data-harmonization-in-the-age-of-machine-learning-a-multicenter-study-across-36-datasets/120652/",4,{"url":52,"name":13,"@type":55,"author":56,"headline":13,"publisher":58,"fileFormat":61,"inLanguage":23,"description":14,"dateModified":62,"datePublished":62,"encodingFormat":61,"isAccessibleForFree":63,"interactionStatistic":64},"DigitalDocument",{"name":9,"@type":57},"Person",{"url":41,"name":59,"@type":60},"DocShare","Organization","application/pdf","2026-08-03",true,{"@type":65,"interactionType":66,"userInteractionCount":20},"InteractionCounter",{"@type":67},"ViewAction",{"@type":69,"mainEntity":70},"FAQPage",[71,77,81],{"name":72,"@type":73,"acceptedAnswer":74},"Why is MRI data harmonization needed in multicenter machine learning studies?","Question",{"text":75,"@type":76},"Because MRI acquisition hardware and protocol parameters introduce non-biological site variability that can confound analyses. Harmonization reduces site effects while aiming to maintain biological content.","Answer",{"name":78,"@type":73,"acceptedAnswer":79},"What problem arises when harmonization is applied to the entire dataset before training?",{"text":80,"@type":76},"Harmonization can lead to data leakage, since information from outside the training set may influence model building. This can falsely overestimate predictive performance.",{"name":82,"@type":73,"acceptedAnswer":83},"What is the proposed harmonizer transformer and how does it help?",{"text":84,"@type":76},"It is an implementation that encapsulates ComBat harmonization within the preprocessing steps of a machine learning pipeline. By fitting harmonization parameters only on training data and applying them to test data, it avoids data leakage.","https://schema.org",{"og:url":52,"og:type":87,"og:title":13,"og:site_name":59,"og:description":14},"article",{"robots":89,"canonical":52},"index,follow",{"doc_id":7,"site_id":24},{"code":4,"msg":5,"data":92},[93,97,101,105,110,115,120,123,128,131,135],{"id":20,"doc_module":4,"doc_module_name":46,"category_name":94,"show_sort_weight":95,"slug":96},"Story & Novel",90,"story-novel",{"id":47,"doc_module":4,"doc_module_name":46,"category_name":98,"show_sort_weight":99,"slug":100},"Literature",80,"literature",{"id":53,"doc_module":4,"doc_module_name":46,"category_name":102,"show_sort_weight":103,"slug":104},"Exam",70,"exam",{"id":106,"doc_module":4,"doc_module_name":46,"category_name":107,"show_sort_weight":108,"slug":109},5,"Comic",60,"comic",{"id":111,"doc_module":4,"doc_module_name":46,"category_name":112,"show_sort_weight":113,"slug":114},6,"Technology",50,"technology",{"id":116,"doc_module":4,"doc_module_name":46,"category_name":117,"show_sort_weight":118,"slug":119},7,"Healthcare",40,"healthcare",{"id":11,"doc_module":4,"doc_module_name":46,"category_name":12,"show_sort_weight":121,"slug":122},30,"research-report",{"id":124,"doc_module":4,"doc_module_name":46,"category_name":125,"show_sort_weight":126,"slug":127},9,"Religion & Spirituality",20,"religion-spirituality",{"id":126,"doc_module":4,"doc_module_name":46,"category_name":129,"show_sort_weight":126,"slug":130},"World Cup","world-cup",{"id":132,"doc_module":4,"doc_module_name":46,"category_name":133,"show_sort_weight":132,"slug":134},10,"Lifestyle","lifestyle",{"id":136,"doc_module":4,"doc_module_name":46,"category_name":137,"show_sort_weight":106,"slug":138},19,"General","general"]