# Metadata Supervised Imaging Representations for Modelling and Controlling Acquisition Variability

arXiv: https://arxiv.org/abs/2607.11295 · HTML: https://mlsanity.com/paper/2607.11295/

## Key facts

- Cited by 1 papers in the map (data through 14 Oct 2026).
- Builds on 8 of its 64 references that are in the map.
- First posted on arXiv 13 Jul 2026.

## Builds on (most cited first, 8 of 8)

- [Representation Learning with Contrastive Predictive Coding](https://mlsanity.com/paper/1807.03748/) (2018)
- [Flow Straight and Fast: Learning to Generate and Transfer Data with Rectified Flow](https://mlsanity.com/paper/2209.03003/) (2022)
- [The RSNA-ASNR-MICCAI BraTS 2021 Benchmark on Brain Tumor Segmentation and Radiogenomic Classification](https://mlsanity.com/paper/2107.02314/) (2021)
- [MONAI: An open-source framework for deep learning in healthcare](https://mlsanity.com/paper/2211.02701/) (2022)
- [Machine Learning with Multi-Site Imaging Data: An Empirical Study on the Impact of Scanner Effects](https://mlsanity.com/paper/1910.04597/) (2019)
- [ContextMRI: Enhancing Compressed Sensing MRI through Metadata Conditioning](https://mlsanity.com/paper/2501.04284/) (2025)
- [DIST-CLIP: Arbitrary Metadata and Image Guided MRI Harmonization via Disentangled Anatomy-Contrast Representations](https://mlsanity.com/paper/2512.07674/) (2025)
- [Understanding Sources of Demographic Predictability in Brain MRI via Disentangling Anatomy and Contrast](https://mlsanity.com/paper/2603.04113/) (2026)

## Built on by (most cited first, 1 of 1)

- [Dataset-Specific Encoding of Image Content Underlies Cross-Dataset Failure in Brain MRI Foundation Models](https://mlsanity.com/paper/2610.13027/) (2026)

## Abstract (arXiv)

Biomedical imaging data exhibit substantial acquisition variability, where identical biological structures can appear markedly different due to differences in imaging devices, acquisition protocols, sites, and reconstruction settings. Consequently, learned representations often entangle underlying biological information with acquisition-dependent appearance, limiting interpretability, generalisation, and clinical deployment. We show that these sources of variation can be disentangled by jointly modelling medical images and acquisition metadata. Using large-scale clinical brain MRI data as a case study, we learn representations that disentangle anatomical structure from contrast-dependent appearance. The resulting framework enables the organisation of heterogeneous imaging protocols, sequence understanding, the detection of image-metadata inconsistencies and imaging artifacts, while preserving biologically relevant anatomical features across diverse acquisitions. Building on these disentangled representations, it further supports generative and translational capabilities, performing both metadata-conditioned synthesis of realistic 3D brain MRIs and anatomy-preserving harmonisation for cross-modality and cross-site adaptation. Our findings demonstrate that acquisition variability is a structured component of the imaging process that can be modeled, audited, synthesised, and controlled, establishing a foundation for acquisition-aware representation learning in large-scale biomedical imaging.