Bernhard Kainz

Semantic Latent Alignment for Flow Matching with Invertible Bi-Lipschitz Adapters

Pretrained latent spaces trade compression, reconstruction, and the learnability of the coordinates a downstream generator has to work in, which complicates adapting these spaces to different domains or tasks: continued training is expensive, and without a new alignment objective it often fails to improve downstream ge...

A Data-Interventional Framework for Auditing Privacy and Fairness in Generative Medical Imaging

Diffusion-based synthetic data generation offers a promising route for sharing medical imaging data without releasing sensitive patient records. However, generative models face a fundamental tension between privacy and fairness: they may memorize rare training samples, leading to privacy risks, or fail to reproduce und...

The Predictability Bias of Joint-Embedding Predictive Learning in Medical Imaging

TAP-JEPA is introduced, a ViT-B/3D trained on 113,611 CT volumes that is competitive with state-of-the-art encoders in scan-level classification and shows that contextual predictability is not a sufficient proxy for clinical relevance in medical representation learning.

Machine Learning for Biomedical Imaging A Data-Interventional Framework for Auditing Privacy and Fairness in Generative Medical Imaging

A Principled Approach to Unsupervised Anomaly Detection

Traditional unsupervised anomaly detection (UAD) methods are designed to flag or localise deviations from a normative distribution, ignoring the underlying generative mechanisms of the anomalies. Yet the nature of an anomaly is often as important as its presence. We reformulate UAD as a Bayesian inverse problem, in whi...

The Text Beside the Image: Detection, Utility and Leakage for Trustworthy Multimodal Medical Data and Beyond

With this ensemble, frequency matching with a public name list recovers zero identities by alignment across the four corpora; the names it got right were ones the detector missed, left in clear text.

MMAP: Multimodal Missing-Aware Pretraining for Longitudinal Alzheimer's Prediction

MMAP is presented, a novel Multimodal Missing-aware Alignment Pretraining method for learning image-tabular representations from incomplete data that outperforms strong multimodal and unimodal baselines on two challenging longitudinal clinical tasks for Alzheimer's disease.

Frozen DINO Localizes Image Edits Without a Localizer

Localized image edits can change a photograph's meaning while leaving most of it authentic, so forensic analysis must identify where an edit occurred. We show that patch-level perturbation responses from frozen DINO encoders are themselves localization maps. Training-free Localization of AI-image Edits from patch-token...