Evaluating normative representation learning in generative AI for robust anomaly detection in brain imaging

Abstract Normative representation learning focuses on understanding the typical anatomical distributions from large datasets of medical scans from healthy individuals. Generative Artificial Intelligence (AI) leverages this attribute to synthesize images that accurately reflect these normative patter...

Full description

Saved in:
Bibliographic Details
Main Authors: Cosmin I. Bercea, Benedikt Wiestler, Daniel Rueckert, Julia A. Schnabel
Format: Article
Language:English
Published: Nature Portfolio 2025-02-01
Series:Nature Communications
Online Access:https://doi.org/10.1038/s41467-025-56321-y
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1850197960875311104
author Cosmin I. Bercea
Benedikt Wiestler
Daniel Rueckert
Julia A. Schnabel
author_facet Cosmin I. Bercea
Benedikt Wiestler
Daniel Rueckert
Julia A. Schnabel
author_sort Cosmin I. Bercea
collection DOAJ
description Abstract Normative representation learning focuses on understanding the typical anatomical distributions from large datasets of medical scans from healthy individuals. Generative Artificial Intelligence (AI) leverages this attribute to synthesize images that accurately reflect these normative patterns. This capability enables the AI allowing them to effectively detect and correct anomalies in new, unseen pathological data without the need for expert labeling. Traditional anomaly detection methods often evaluate the anomaly detection performance, overlooking the crucial role of normative learning. In our analysis, we introduce novel metrics, specifically designed to evaluate this facet in AI models. We apply these metrics across various generative AI frameworks, including advanced diffusion models, and rigorously test them against complex and diverse brain pathologies. In addition, we conduct a large multi-reader study to compare these metrics to experts’ evaluations. Our analysis demonstrates that models proficient in normative learning exhibit exceptional versatility, adeptly detecting a wide range of unseen medical conditions. Our code is available at https://github.com/compai-lab/2024-ncomms-bercea.git .
format Article
id doaj-art-3e51bfb18ec749f28155672166d0f3d5
institution OA Journals
issn 2041-1723
language English
publishDate 2025-02-01
publisher Nature Portfolio
record_format Article
series Nature Communications
spelling doaj-art-3e51bfb18ec749f28155672166d0f3d52025-08-20T02:12:59ZengNature PortfolioNature Communications2041-17232025-02-0116111010.1038/s41467-025-56321-yEvaluating normative representation learning in generative AI for robust anomaly detection in brain imagingCosmin I. Bercea0Benedikt Wiestler1Daniel Rueckert2Julia A. Schnabel3Chair of Computational Imaging and AI in Medicine, Technical University of Munich (TUM)Chair of AI for Image-Guided Diagnosis and Therapy, TUM School of Medicine and HealthMunich Center for Machine Learning (MCML)Chair of Computational Imaging and AI in Medicine, Technical University of Munich (TUM)Abstract Normative representation learning focuses on understanding the typical anatomical distributions from large datasets of medical scans from healthy individuals. Generative Artificial Intelligence (AI) leverages this attribute to synthesize images that accurately reflect these normative patterns. This capability enables the AI allowing them to effectively detect and correct anomalies in new, unseen pathological data without the need for expert labeling. Traditional anomaly detection methods often evaluate the anomaly detection performance, overlooking the crucial role of normative learning. In our analysis, we introduce novel metrics, specifically designed to evaluate this facet in AI models. We apply these metrics across various generative AI frameworks, including advanced diffusion models, and rigorously test them against complex and diverse brain pathologies. In addition, we conduct a large multi-reader study to compare these metrics to experts’ evaluations. Our analysis demonstrates that models proficient in normative learning exhibit exceptional versatility, adeptly detecting a wide range of unseen medical conditions. Our code is available at https://github.com/compai-lab/2024-ncomms-bercea.git .https://doi.org/10.1038/s41467-025-56321-y
spellingShingle Cosmin I. Bercea
Benedikt Wiestler
Daniel Rueckert
Julia A. Schnabel
Evaluating normative representation learning in generative AI for robust anomaly detection in brain imaging
Nature Communications
title Evaluating normative representation learning in generative AI for robust anomaly detection in brain imaging
title_full Evaluating normative representation learning in generative AI for robust anomaly detection in brain imaging
title_fullStr Evaluating normative representation learning in generative AI for robust anomaly detection in brain imaging
title_full_unstemmed Evaluating normative representation learning in generative AI for robust anomaly detection in brain imaging
title_short Evaluating normative representation learning in generative AI for robust anomaly detection in brain imaging
title_sort evaluating normative representation learning in generative ai for robust anomaly detection in brain imaging
url https://doi.org/10.1038/s41467-025-56321-y
work_keys_str_mv AT cosminibercea evaluatingnormativerepresentationlearningingenerativeaiforrobustanomalydetectioninbrainimaging
AT benediktwiestler evaluatingnormativerepresentationlearningingenerativeaiforrobustanomalydetectioninbrainimaging
AT danielrueckert evaluatingnormativerepresentationlearningingenerativeaiforrobustanomalydetectioninbrainimaging
AT juliaaschnabel evaluatingnormativerepresentationlearningingenerativeaiforrobustanomalydetectioninbrainimaging