TUM Klinikum

Trained on 909,873 2D CT and MRI slices from the TotalSegmentator dataset, this study fits Matryoshka sparse autoencoders with BatchTopK sparsification on frozen embeddings from BiomedParse (biomedical) and DINOv3 (general-purpose) foundation models alongside a random-weight baseline, finding that sparse features reconstruct original embeddings with R2 up to 0.941, recover 87.8% of downstream performance with only 10 features (99.4% dimensionality reduction), preserve 97.7% of dense retrieval quality with five-feature fingerprints, correspond to monosemantic concepts expressible in language as verified by an independent LLM judge, and enable zero-shot language-driven image retrieval on a single clinical text query.

Trained on 909,873 2D CT and MRI slices from the TotalSegmentator dataset, this study fits Matryoshka sparse autoencoders with BatchTopK sparsification on frozen embeddings from BiomedParse (biomedical) and DINOv3 (general-purpose) foundation models alongside a random-weight baseline, finding that sparse features reconstruct original embeddings with R2 up to 0.941, recover 87.8% of downstream performance with only 10 features (99.4% dimensionality reduction), preserve 97.7% of dense retrieval quality with five-feature fingerprints, correspond to monosemantic concepts expressible in language as verified by an independent LLM judge, and enable zero-shot language-driven image retrieval on a single clinical text query.

The work introduces ConRad, a reinforcement learning framework that fine-tunes medical large vision-language models with the GRPO algorithm and a logarithmic scoring-rule reward so they emit verbalized confidence alongside radiology reports at report or sentence level, reducing report-level ECE to 0.106 and sentence-level ECE to 0.239 on MIMIC-CXR and improving zero-shot IU-Xray ECE from 0.310 to 0.034 while keeping report quality (GREEN) stable.

The work introduces ConRad, a reinforcement learning framework that fine-tunes medical large vision-language models with the GRPO algorithm and a logarithmic scoring-rule reward so they emit verbalized confidence alongside radiology reports at report or sentence level, reducing report-level ECE to 0.106 and sentence-level ECE to 0.239 on MIMIC-CXR and improving zero-shot IU-Xray ECE from 0.310 to 0.034 while keeping report quality (GREEN) stable.

The work proposes Gabor primitives for MRI reconstruction, modulating each Gaussian envelope with a complex exponential so its spectral support can be placed at an arbitrary k-space location, and designs a two-basis low-rank temporal model separating geometry and signal-intensity dynamics; on 99 Cartesian (R=12, R=16) and 102 radial (R≈23) cardiac cine acquisitions, Gabor primitives achieve the highest PSNR and SSIM in all settings, improving over Gaussian primitives by +1.11/+0.72/+0.86 dB and over PICS by +2.34 dB on radial data, with a parameter ratio ρ<0.5 and continuous-resolution evaluation enabling 4× super-resolution.

The work proposes Gabor primitives for MRI reconstruction, modulating each Gaussian envelope with a complex exponential so its spectral support can be placed at an arbitrary k-space location, and designs a two-basis low-rank temporal model separating geometry and signal-intensity dynamics; on 99 Cartesian (R=12, R=16) and 102 radial (R≈23) cardiac cine acquisitions, Gabor primitives achieve the highest PSNR and SSIM in all settings, improving over Gaussian primitives by +1.11/+0.72/+0.86 dB and over PICS by +2.34 dB on radial data, with a parameter ratio ρ<0.5 and continuous-resolution evaluation enabling 4× super-resolution.