arXiv:2607. 07219v1 Announce Type: cross Abstract: Vision foundation models (VFMs) are increasingly being developed for radiological imaging, yet their definition, development and evaluation remain heterogeneous.
arXiv:2601.15891v4 Announce Type: replace Abstract: Vision-language pretraining has driven progress in medical image representation learning, but it depends on paired image-text data and can inherit...
nnFoundation introduces complementary convolutional and transformer-based 3D foundation models for radiology, trained on 2.1 million CT, MRI, and PET volumes from 125 datasets. The models are evaluated on 108 tasks—including segmentation, detection, classification, report generation, and image retrieval—under domain shift, low-data, and low-compute scenarios, consistently outperforming prior 3D foundation models and training from scratch. Performance varies by task type, with convolutional models excelling at spatially localized tasks and transformer models at global semantic reasoning, and dynamic alignment with dataset characteristics further enhances transferability.
arXiv:2409.16183v2 Announce Type: replace Abstract: Radiology is a vital and complex component of modern clinical workflow and covers many tasks. Recently, vision-language (VL) foundation models in m...
The paper investigates how medical vision‑language models (VLMs) behave when faced with distribution shifts such as changes in acquisition domain, supervision, or evaluation protocol. Using datasets like NIH ChestXray14, CheXpert, PadChest, and OpenI, the authors isolate cross‑dataset visual transfer, evaluate multimodal alignment, and quantify source‑proxy leakage in frozen embeddings. They find that self‑supervised visual initialization improves transfer, adversarial adaptation is only marginally helpful, and that multimodal retrieval performance drops under external stress tests while source‑proxy information remains recoverable, highlighting hidden failure modes in medical VLMs.
arXiv:2608.21571v1 Announce Type: new Abstract: Lung cancer remains a leading cause of cancer-related mortality worldwide, and early diagnosis is critical for improving survival. However, early-stage...
arXiv:2608.28773v1 Announce Type: cross Abstract: Perceptual losses are widely used in medical image synthesis because they encourage agreement in high-level structure beyond voxel-wise intensity sim...
arXiv:2607. 17782v1 Announce Type: cross Abstract: Foundation models pretrained using self-supervised learning have transformed computer vision by learning transferable representations from large-scale unlabeled data.
The paper presents the first systematic evaluation of out‑of‑distribution generalization for congenital heart disease (CHD) segmentation, using the ImageCHD cohort as a held‑out target. It compares several segmentation architectures under different training regimes, showing that in‑distribution performance is a poor predictor of cross‑cohort robustness: nnU‑Net drops from 0.77 to 0.51 Dice, while SwinUNETR maintains higher performance at 0.67 Dice. Limited target‑domain adaptation with only 11 labeled ImageCHD cases boosts all SwinUNETR variants above 0.76 Dice, highlighting the importance of explicit cross‑dataset testing.
arXiv:2605. 23995v4 Announce Type: replace-cross Abstract: Self-supervised learning (SSL) is increasingly used in medical image analysis to reduce dependence on costly expert annotations by learning transferable representations from unlabeled data.
The paper explores large‑scale pretraining to enhance deep learning‑based geometric distortion correction for diffusion‑weighted imaging (DWI). By framing the task as image reconstruction, the authors compare a non‑pretrained baseline with self‑supervised and generative pretrained models, finding that the cWDM model yields the best quantitative and qualitative results. When applied to low‑resource, high‑throughput settings in a low‑ and middle‑income country, the pretrained models faced transferability issues, but aligning images to a common standard space improved predictions, indicating that harmonized preprocessing can aid cross‑domain deployment.
CMRVision is a cardiac magnetic resonance (CMR) foundation model trained with DINOv3-style self‑supervised learning on 36 million multi‑center, multi‑sequence CMR images. It outperforms prior natural‑image, medical‑image, supervised, and CMR baselines on multi‑task segmentation (cine, LGE, mapping) and cine view classification, achieving Dice scores of 0.940–0.967 for LV and 0.855–0.905 for myocardium, and a zero‑shot Dice of 0.692 on unseen LGE long‑axis views. The model demonstrates robust cross‑view generalization and highest average accuracy (0.906) for cine view classification.
arXiv:2605. 23995v2 Announce Type: replace-cross Abstract: Self-supervised learning (SSL) has emerged as a promising paradigm for addressing the annotation bottleneck in medical imaging by learning representations from unlabeled data.