Challenges and best practices when using ComBAT to harmonize diffusion MRI data
作者:Pierre‐Marc Jodoin, Manon Edde, Gabriel Girard, Félix Dumais, Guillaume Theaud, Matthieu Dumont, Jean-Christophe Houde, Yoan David, Maxime Descoteaux, Alzheimer’s Disease Neuroimaging Initiative, Michael W. Weiner, Paul Aisen, Ronald Petersen, Clifford R. Jack, William Jagust, Susan Landau, Monica Rivera-Mindt, Ozioma Okonkwo, Leslie M. Shaw, Edward B. Lee, Arthur W. Toga, Laurel Beckett, Danielle Harvey, Robert C. Green, Andrew J. Saykin, Kwangsik Nho, Richard J. Perrin, Duygu Tosun, Pallavi Sachdev, Erin Drake, Tom Montine, Cat Conti, Rachel L. Nosheny, Diana Truran‐Sacrey, Juliet Fockler, Melanie J. Miller, Catherine Conti, Winnie Kwang, Chengshi Jin, Adam Diaz, Miriam T. Ashford, Derek Flenniken, Adrienne Kormos, Michael Rafii, Rema Raman, Gustavo Jiménez, Michael Donohue, Jennifer Salazar, Andrea Fidell, Virginia Boatwright, Justin Robison, Caileigh Zimmerman, Yuliana Cabrera, Sarah Walter, Taylor Clanton, Elizabeth Shaffer, Caitlin Webb, Lindsey Hergesheimer, Stephanie R. Rainey‐Smith, Sheila Ogwang, Olusegun Adegoke, Payam Mahboubi, Jeremy Pizzola, Cecily Jenkins, Joel P. Felmlee, Nick C. Fox, Paul M. Thompson, Charles DeCarli, Arvin Forghanian-Arani, Bret Borowski, Calvin Reyes, Caitie Hedberg, Chad Ward, Christopher G. Schwarz, Denise Reyes, Jeff Gunter, John Moore-Weiss, Kejal Kantarci, Leonard Matoush, Matthew L. Senjem, Prashanthi Vemuri, Robert W. Reid, Ian G. Malone, Sophia I. Thomopoulos, Talia M. Nir, Neda Jahanshad, Alexander Knaack, Evan Fletcher, Duygu Tosun, Stephanie Rossi Chen, Mark Choe, Karen Crawford, Paul A. Yushkevich, Sandhitsu R. Das, Laurel Beckett, Naomi Saito, Kedir Adem Hussen, Ozioma Okonkwo, Hannatu Amaza, Mai Seng Thao, Matt Glittenberg, Isabella Hoang, Joe Strong, Trinity Weisensel, Fabiola Magana, Lisa Thomas, Kaori Kubo Germano, Sandra Talavera, Vanessa Guzmán, Adeyinka Ajayi, Joseph Di Benedetto, Shaniya Parkins, Omobolanle Ayo, Victor L. Villemagne, Brian J. Lopresti, Robert A. Koeppe, Gil D. Rabinovici, John C. Morris, Erin Franklin, Nigel J. Cairns, Lisa Taylor‐Reinwald, Virginia M.‐Y. Lee, Magdalena Korecka, Magdalena Brylska, Yang Wan, John Q. Trojanowki, Scott Neu, Tatiana Foroud, Taeho Jo, Shannon L. Risacher, Hannah Craft, Liana G. Apostolova, Kelly Nudelman, Kelley Faber, ZoA Potter, Kaci Lacy, Rima Kaddurah-Daouk, Li Shen, David N. Soleimani‐Meigooni, Renaud La Joie, Konstantinos Chiotis, Maison Abu Raya, Agathe Vrillon, Charles Windon, Julien Lagarde, Zoe Lin, Aidyn Rose Hills, Jason Karlawish, Emily A. Largent, Kristin Harkins, Joshua D. Grill, Zaven Kachaturian, R.T. Frank, Peter J. Snyder, Neil Buckholtz, John Hsiao, Laurie Ryan, Susan Molchan, Marı́a C. Carrillo, William Z. Potter, Lisa Barnes, Héctor Alfredo Baptista González, Carole Ho, Jonathan Jackson, Eliezer Masliah, Donna Masterman, Nina Silverberg · 发表于:Scientific Reports · 年份:2025 · DOI:10.1038/s41598-025-25400-x · 被引用次数:6 · 研究领域:Advanced Neuroimaging Techniques and Applications、Functional Brain Connectivity Studies、MRI in cancer diagnosis
Over the years, ComBAT has become the standard method for harmonizing MRI-derived measurements, with its ability to compensate for site-related additive and multiplicative biases while preserving biological variability. However, ComBAT relies on a set of assumptions that, when violated, can result in flawed harmonization. In this paper, we thoroughly review ComBAT's mathematical foundation, outlining these assumptions, and exploring their implications for the demographic composition necessary for optimal results. Through a series of experiments involving a slightly modified version of ComBAT called Pairwise-ComBAT tailored for normative modeling applications, we assess the impact of various population characteristics, including population size, age distribution, the absence of certain covariates, and the magnitude of additive and multiplicative factors. Based on these experiments, we present five essential recommendations that should be carefully considered to enhance consistency and supporting reproducibility, two essential factors for open science, collaborative research, and real-life clinical deployment.