Health AI Central
Peer-Reviewed Publication
Med Image Anal2026;110103943.May 1, 2026Journal Article

Fréchet radiomic distance (FRD): A versatile metric for comparing medical imaging datasets.

Nicholas Konz1, Richard Osuala2, Preeti Verma3, Yuwen Chen4, Hanxue Gu5, Haoyu Dong6, Yaqian Chen7, Andrew Marshall8, Lidia Garrucho9, Kaisar Kushibar10, Daniel M Lang11, Gene S Kim12, Lars J Grimm13, John M Lewin14, James S Duncan15, Julia A Schnabel16, Oliver Diaz17, Karim Lekadir18, Maciej A Mazurowski19
1Department of Electrical and Computer Engineering, Duke University, NC, USA. Electronic address: nicholas.konz@duke.edu.
2Departament de Matemàtiques i Informàtica, Universitat de Barcelona, Barcelona, Spain; Institute of Machine Learning in Biomedical Imaging, Helmholtz Center Munich, Munich, Germany; School of Computation, Information and Technology, Technical University of Munich, Munich, Germany. Electronic address: richard.osuala@ub.edu.
3Departament de Matemàtiques i Informàtica, Universitat de Barcelona, Barcelona, Spain.
4Department of Electrical and Computer Engineering, Duke University, NC, USA. Electronic address: yuwen.chen@duke.edu.
5Department of Electrical and Computer Engineering, Duke University, NC, USA. Electronic address: hanxue.gu@duke.edu.
6Department of Electrical and Computer Engineering, Duke University, NC, USA. Electronic address: haoyu.dong151@duke.edu.
7Department of Electrical and Computer Engineering, Duke University, NC, USA. Electronic address: yaqian.chen@duke.edu.
8Department of Biomedical Engineering, Yale University, CT, USA. Electronic address: andrew.marshall@yale.edu.
9Departament de Matemàtiques i Informàtica, Universitat de Barcelona, Barcelona, Spain. Electronic address: lgarrucho@ub.edu.
10Departament de Matemàtiques i Informàtica, Universitat de Barcelona, Barcelona, Spain. Electronic address: kaisar.kushibar@ub.edu.
11Institute of Machine Learning in Biomedical Imaging, Helmholtz Center Munich, Munich, Germany; School of Computation, Information and Technology, Technical University of Munich, Munich, Germany. Electronic address: lang@helmholtz-munich.de.
12Department of Radiology, Weill Cornell Medical College, NY, USA. Electronic address: sgk4001@med.cornell.edu.
13Department of Radiology, Duke University, NC, USA. Electronic address: lars.grimm@duke.edu.
14Department of Radiology & Biomedical Imaging, Yale University, CT, USA. Electronic address: john.lewin@yale.edu.
15Department of Biomedical Engineering, Yale University, CT, USA; Department of Radiology & Biomedical Imaging, Yale University, CT, USA; Department of Electrical Engineering, Yale University, CT, USA. Electronic address: james.duncan@yale.edu.
16Institute of Machine Learning in Biomedical Imaging, Helmholtz Center Munich, Munich, Germany; School of Computation, Information and Technology, Technical University of Munich, Munich, Germany; School of Biomedical Engineering & Imaging Sciences, King's College London, London, UK. Electronic address: julia.schnabel@tum.de.
17Departament de Matemàtiques i Informàtica, Universitat de Barcelona, Barcelona, Spain; Computer Vision Center, Universitat Autònoma de Barcelona, Bellaterra, Spain. Electronic address: oliver.diaz@ub.edu.
18Departament de Matemàtiques i Informàtica, Universitat de Barcelona, Barcelona, Spain; Institució Catalana de Recerca i Estudis Avançats (ICREA), Barcelona, Spain. Electronic address: karim.lekadir@ub.edu.
19Department of Electrical and Computer Engineering, Duke University, NC, USA; Department of Radiology, Duke University, NC, USA; Department of Biostatistics & Bioinformatics, Duke University, NC, USA; Department of Computer Science, Duke University, NC, USA. Electronic address: maciej.mazurowski@duke.edu.
PubMed PMC DOI PDF

Abstract

Determining whether two sets of images belong to the same or different distributions or domains is a crucial task in modern medical image analysis and deep learning; for example, to evaluate the output quality of image generative models. Currently, metrics used for this task either rely on the (potentially biased) choice of some downstream task, such as segmentation, or adopt task-independent perc…

Create a free account to keep reading

Free members get 10 full research views every month across publications, clinical trials, FDA clearances, adverse events, and NIH grants. No credit card required.

Create Free AccountSign In

Want unlimited research access? See Pro plans

Data Accuracy Notice: Research intelligence on Health AI Central is aggregated from public sources (PubMed, ClinicalTrials.gov, FDA, NIH, CMS, and others) and refreshed nightly. Classifications and derived metrics are produced by automated methods described in our Methodology. We recommend verifying critical data points against the primary sources before making decisions.