Image-conditioned latent rectified flow models for 3D medical anomaly localisation.

Overview

1. Biomedical Image Analysis Group, Department of Computing, Imperial College London, London, United Kingdom
2. IDEA Lab, Department of Artificial Intelligence in Biomedical Engineering, Friedrich-Alexander Universität Erlangen-Nürnberg, Erlangen, Germany
3. Structured and Probabilistic Intelligent Knowledge Engineering Group, Department of Computing, Imperial College London, London, United Kingdom

Abstract

Introduction: Reconstruction-based methods offer a promising solution for unsupervised anomaly detection in medical imaging tasks. These methods train generative models on healthy data alone and identify anomalies as deviations between an input image and its pseudo-healthy reconstruction. The downfall of these methods is their dependence on two assumptions that often fail in practice: that models cannot reproduce unseen pathologies, yet can faithfully reconstruct healthy tissue. A recent image-conditioned diffusion approach explicitly addresses these issues by training a model to restore synthetic anomalies inserted into healthy images. However, it operates in 2D pixel space, discarding inter-slice context and incurring high computational cost.

Methods: We address both limitations by performing image-conditioned restoration in a 3D latent space using a pretrained VAE and rectified flow, capturing volumetric context whilst drastically reducing computational overhead. To mitigate false positives introduced by VAE compression, we propose using the restoration change which measures the difference between the pseudo-healthy latent restoration and the VAE reconstruction of the original, rather than the standard reconstruction error. We further experiment with applying the synthetic anomaly training task directly in latent space to improve sensitivity to low-contrast anomalies.

Results: We perform extensive experiments across various medical imaging benchmarks, including brain MRI and the newly released AADD dataset, comparing against reconstruction-based, feature-modelling, attention-based and self-supervised anomaly detection methods. Our image-conditioned rectified flow models establish a new state-of-the-art, with an ensemble of models trained with pixel-space and latent-space anomalies yielding the strongest overall performance.

Discussion: These results demonstrate how incorporating 3D context enables better anomaly detection performance whilst also being ∼10 times faster. The difference in performance between models trained using latent-space and pixel-space anomalies suggests that further broadening of the anomaly imputation process could continue to improve the robustness of these models. Such improvements are certainly necessary, as the AADD benchmark is far from being saturated. Code is available at https://github.com/matt-baugh/img-cond-latent-rflow-model-ad.

Reproduced under the paper's license (CC BY), from the paper cited above.

Code

The paper links to its data, not to its authors' code: see the Data section.

Tracing map

A tracing map links a paper to the code its authors published: this paper has none, so it has no map.

Data

Datasets cited

• huggingface.co/datasets/jemtan11/aadd, at Hugging Face; found in “Data availability statement”
• kaggle.com/datasets/awsaf49, at Kaggle; found in “Data availability statement”

Data availability statement

Publicly available datasets were analyzed in this study. This data can be found here: Advancing Anomaly Detection Dataset (AADD): https://huggingface.co/datasets/Jemtan11/AADD; CamCan: https://camcan-archive.mrc-cbu.cam.ac.uk/dataaccess/index.php; HCP: https://balsa.wustl.edu/project?project=HCP_YA; IXI: https://brain-development.org/ixi-dataset/; BraTs: https://www.kaggle.com/datasets/awsaf49/brats20-dataset-training-validation; ATLASv2: https://fcon_1000.projects.nitrc.org/indi/retro/atlas.html.

Reproduced under the paper's license (CC BY), from the paper cited above.

Versions

The history of this record: each version stored by the harvester or made by a correction of its authors or of the maintainers of its code, and what changed in its facts. The texts of the paper (its abstract, its availability statements) are not part of it; versions that changed only those are not listed.

Version 1, 27 September 2026: the first record

Recorded: type, language, journal, volume, pages, dates, 7 authors, 6 keywords, 1 funder, 13 references.

Cite

This paper

Baugh, M., Müller, J. P., Cechnicka, S., Gu Baugh, K., Bonnici, J., Myles, J., & Kainz, B. (2026). Image-conditioned latent rectified flow models for 3D medical anomaly localisation. Frontiers in radiology, 6, 1847193. https://doi.org/10.3389/fradi.2026.1847193

Similar papers

The papers with a page that share the most with this one: the tools found in their code, their categories, datasets, cited references and authors, the rarest counting most.

[1] doi:10.3389/frai.2026.1807248 [code]

Federated learning framework for medical image analysis with perspective-aware contrastive and mixture of experts.

Journal: Frontiers in artificial intelligence

In common: kaggle.com/datasets/awsaf49, methods / tools, 2 references

[2] doi:10.3390/jimaging12060251

3D Deep Learning for Brain Tumor Segmentation and Survival Prediction: A Comprehensive Multi-Modal Analysis Using the BraTS2020 Dataset.

Journal: Journal of imaging

In common: kaggle.com/datasets/awsaf49, 2 references

[3] doi:10.1038/s41598-026-53337-2 [code]

Progression-guided spatiotemporal memory transformers for accurate and consistent longitudinal brain tumor segmentation.

Journal: Scientific reports

In common: kaggle.com/datasets/awsaf49, methods / tools, 1 reference

[4] doi:10.3389/frai.2026.1812932

Bridging global context and local precision using a disagreement-based region specific ensemble of Swin UNETR and SegResNet for 3D glioma segmentation.

Journal: Frontiers in artificial intelligence

In common: kaggle.com/datasets/awsaf49, methods / tools

[5] doi:10.1371/journal.pone.0351667

MamNet-PT: A Mamba-enhanced hybrid architecture with selective state-space modeling for uncertainty-aware brain tumor segmentation.

Journal: PloS one

In common: kaggle.com/datasets/awsaf49, methods / tools

[6] doi:10.1038/s41598-026-50240-8

High-precision brain tumor segmentation with switchable normalization in faster R-CNN architecture.

Journal: Scientific reports

In common: kaggle.com/datasets/awsaf49, methods / tools

[7] doi:10.1038/s41598-026-45697-6

Real-time medical images enhancement technique using Adaptive Histogram Equalization model.

Journal: Scientific reports

In common: kaggle.com/datasets/awsaf49, methods / tools

[8] doi:10.1186/s12911-026-03551-9

A multi-sequence MRI integration framework using SwinUNETR-v2 for multiple sclerosis lesion segmentation.

Journal: BMC medical informatics and decision making

In common: kaggle.com/datasets/awsaf49

[9] doi:10.1038/s41467-026-71267-5 [code]

Human-like cognitive generalization for large models via mental representation-guided supervision.

Journal: Nature communications

In common: kaggle.com/datasets/awsaf49

[10] doi:10.3389/fonc.2026.1928589

LiteFreqMamba:lightweight frequency-enhanced Mamba for efficient 3D brain tumor segmentation.

Journal: Frontiers in oncology

In common: methods / tools, 2 references

Contribute

The authors of this paper can claim it, correct its record and validate its tracing map, and the maintainers of its code (its owner, or a public member of its organization) correct what it says of their repository; anyone signed in can ask for its removal. Every request goes to OSCR's own machine, which answers it; your account page follows them.

Request its removal

To ask OSCR to remove this record, the copies of its authors' scripts or its tracing map, use the removal request page: signed in, you say who you are, what to remove and why, then review and confirm the request. Published rules decide every request (how).

Discussion: questions and error reports about this paper and its code, from signed-in readers and its authors. It opens with sign-in.

Reproductions: reports from readers who ran the authors' code: what they reproduced, with which environment, commit and data. It opens with sign-in.

Activity: what happens around this paper: new versions of its record, its map's validation, discussions and reproductions. It opens with sign-in.