Paul Friedrich

PhD Thesis Defended & Best Paper Award at ShapeMI 2026

I successfully defended my PhD thesis Efficient Medical Image Processing Across Representations at the University of Basel and graduated summa cum laude.

In addition, our paper AutoFFS: Adversarial Deformations for Facial Feminization Surgery Planning won the Best Paper Award at the Workshop on Shape in Medical Imaging (ShapeMI @ MICCAI 2026).

Paper Accepted at Nature Scientific Reports

Our paper Deep generative models for enhanced vitreous OCT imaging has been accepted at Nature Scientific Reports.

2 Papers Accepted at MICCAI 2026 Workshops

I (co-)authored 2 papers accepted at MICCAI workshops. Make sure to check out:

• AutoFFS: Adversarial Deformations for Facial Feminization Surgery Planning accepted at ShapeMI
• Modelling Geographic Atrophy Progression using Implicit Neural Representations accepted at Off-Grid

Paper accepted at ICML 2026

Our paper Optimizing Rank for High-Fidelity Implicit Neural Representations has been accepted at ICML 2026. We show that the low-frequency bias of vanilla MLP-based INRs stems from rank degradation during training rather than inherent architectural limitations, and that preserving high-rank updates, using optimizers such as MUON, consistently improves INR performance across network architectures.

Paper accepted at MIDL 2026 - Oral

Our paper MedFuncta: A Unified Framework for Learning Efficient Medical Neural Fields has been accepted as an oral at MIDL 2026. We introduce a modality agnostic continuous data representation based on meta-learned neural fields (NFs) and show how NFs can be generalized from single instances to large datasets.

Paper accepted at ICLR 2026

Our paper Beyond Uniformity: Regularizing Implicit Neural Representations Through a Lipschitz Lens has been accepted at ICLR 2026. We frame Lipschitz regularization as a budget allocation problem, investigating the performance of a diverse set of strategies on different inverse problems.