King's College London

Research portal

Neuroharmony: A new tool for harmonizing volumetric MRI data from unseen scanners

Research output: Contribution to journalArticle

Rafael Garcia Dias, Cristina Scarpazza, Lea Baecker, Sandra Mendes Vieira, Walter Lopez Pinaya, Aiden Corvin, Alberto Redolf, Barnaby Nelson, Benedicto Crespo-Facorro, Colm McDonald, Diana Tordesillas-Gutiérrez, Dara Cannon, David Mothersill, Dennis Hernaus, Derek Morris, Esther Setien-Suero, Gary Donohoe, Giovanni Frisoniq, Giulia Tronchin, Joao Sato & 15 more Machteld Marcelis, Matthew Kempton, Neeltje E. M. van Haren, Oliver Gruber, Patrick McGorry, Paul Amminger, Philip McGuire, Qiyong Gong, René S. Kahnz, Rosa Ayesa-Arriola, Therese van Amelsvoort, Victor Ortiz-Garcia de la Foz, Vince Calhoun, Wiepke Cahn, Andrea Mechelli

Original languageEnglish
JournalNeuroImage
Publication statusAccepted/In press - 30 Jun 2020

King's Authors

Abstract

The increasing availability of magnetic resonance imaging (MRI) datasets is boosting the interest in the application ofmachine learning in neuroimaging. A key challenge to the development of reliable machine learning models, and theirtranslational implementation in real-word clinical practice, is the integration of datasets collected using differentscanners. Current approaches for harmonizing multi-scanner data, such as the ComBat method, require a statisticallyrepresentative sample; therefore, these approaches are not suitable for machine learning models aimed at clinicaltranslation where the focus is on the assessment of individual scans from previously unseen scanners. To overcome thischallenge, we developed a tool (‘Neuroharmony’) that is capable of harmonizing single images from unseen/unknownscanners based on a set of image quality metrics, i.e. intrinsic characteristics which can be extracted from individual images without requiring a statistically representative sample. The tool was developed using a mega-dataset ofneuroanatomical data from 15,026 healthy subjects to train a machine learning model that captures the relationshipbetween image quality metrics and the relative volume corrections for each region of the brain prescribed by theComBat method. The tool resulted to be effective in reducing systematic scanner-related bias from new individualimages taken from unseen scanners without requiring any specifications about the image acquisition. Our approachrepresents a significant step forward in the quest to develop reliable imaging-based clinical tools. Neuroharmony andthe instructions on how to use it are available at https://github.com/garciadias/Neuroharmony.

View graph of relations

© 2018 King's College London | Strand | London WC2R 2LS | England | United Kingdom | Tel +44 (0)20 7836 5454