Cosmin Ciausu

Leonard Nürnberg, Dennis Bontempi, Suraj Pai et al.

Artificial intelligence (AI) has the potential to transform medical imaging by automating image analysis and accelerating clinical research. However, research and clinical use are limited by the wide variety of AI implementations and architectures, inconsistent documentation, and reproducibility issues. Here, we introduce MHub.ai, an open-source, container-based platform that standardizes access to AI models with minimal configuration, promoting accessibility and reproducibility in medical imaging. MHub.ai packages models from peer-reviewed publications into standardized containers that support direct processing of DICOM and other formats, provide a unified application interface, and embed structured metadata. Each model is accompanied by publicly available reference data that can be used to confirm model operation. MHub.ai includes an initial set of state-of-the-art segmentation, prediction, and feature extraction models for different modalities. The modular framework enables adaptation of any model and supports community contributions. We demonstrate the utility of the platform in a clinical use case through comparative evaluation of lung segmentation models. To further strengthen transparency and reproducibility, we publicly release the generated segmentations and evaluation metrics and provide interactive dashboards that allow readers to inspect individual cases and reproduce or extend our analysis. By simplifying model use, MHub.ai enables side-by-side benchmarking with identical execution commands and standardized outputs, and lowers the barrier to clinical translation.

Suraj Pai, Thibault Heintz, Cosmin Ciausu et al.

Zero-shot vision-language models (VLMs) offer a promptable alternative to task-specific training for gross tumor volume (GTV) delineation in non-small-cell lung cancer (NSCLC), but the prompt dimensions that govern their spatial behavior remain poorly understood. We study this question by probing alignment directions in VoxTell on a held-out internal NSCLC tumor dataset through sub-prompt decomposition into diagnosis, demographic, staging, anatomical, generic, and irrelevant controls; attribute-wise perturbation robustness; specificity ladders; and cross-case prompt swaps, while benchmarking against fine-tuned and zero-shot baselines using the Dice Similarity Coefficient (DSC) with Wilcoxon signed-rank tests and Benjamini-Hochberg correction. Alignment analyses revealed that anatomical location is the dominant driver of VoxTell's spatial attention: 63.4 percent of location perturbations caused catastrophic drops, prompt specificity improved from generic to full descriptions except for diagnosis-only prompts, irrelevant prompts correctly yielded zero segmentation, and cross-case prompt swaps confirmed patient-specific conditioning (matched DSC 0.906 vs. mismatched 0.406). Histology and stage substitutions had minimal effect, indicating that the model prioritizes "where to look" over "what to look for." In this context, VoxTell, operating fully zero-shot, achieved a mean DSC of 0.613, statistically indistinguishable from nnUNet (0.690, adjusted p = 0.156) and Ahmed et al. (0.675, adjusted p = 0.679), while significantly outperforming all other zero-shot models. Together, these findings argue that segmentation VLMs should be evaluated not only by Dice, but also by the prompt dimensions to which they align.

Cosmin Ciausu, Deepa Krishnaswamy, Benjamin Billot et al.

Deep learning has shown great promise in the ability to automatically annotate organs in magnetic resonance imaging (MRI) scans, for example, of the brain. However, despite advancements in the field, the ability to accurately segment abdominal organs remains difficult across MR. In part, this may be explained by the much greater variability in image appearance and severely limited availability of training labels. The inherent nature of computed tomography (CT) scans makes it easier to annotate, resulting in a larger availability of expert annotations for the latter. We leverage a modality-agnostic domain randomization approach, utilizing CT label maps to generate synthetic images on-the-fly during training, further used to train a U-Net segmentation network for abdominal organs segmentation. Our approach shows comparable results compared to fully-supervised segmentation methods trained on MR data. Our method results in Dice scores of 0.90 (0.08) and 0.91 (0.08) for the right and left kidney respectively, compared to a pretrained nnU-Net model yielding 0.87 (0.20) and 0.91 (0.03). We will make our code publicly available.

Deepa Krishnaswamy, Cosmin Ciausu, Steve Pieper et al.

Recent advances in deep learning have led to robust automated tools for segmentation of abdominal computed tomography (CT). Meanwhile, segmentation of magnetic resonance imaging (MRI) is substantially more challenging due to the inherent signal variability and the increased effort required for annotating training datasets. Hence, existing approaches are trained on limited sets of MRI sequences, which might limit their generalizability. To characterize the landscape of MRI abdominal segmentation tools, we present here a comprehensive benchmarking of the three state-of-the-art and open-source models: MRSegmentator, MRISegmentator-Abdomen, and TotalSegmentator MRI. Since these models are trained using labor-intensive manual annotation cycles, we also introduce and evaluate ABDSynth, a SynthSeg-based model purely trained on widely available CT segmentations (no real images). More generally, we assess accuracy and generalizability by leveraging three public datasets (not seen by any of the evaluated methods during their training), which span all major manufacturers, five MRI sequences, as well as a variety of subject conditions, voxel resolutions, and fields-of-view. Our results reveal that MRSegmentator achieves the best performance and is most generalizable. In contrast, ABDSynth yields slightly less accurate results, but its relaxed requirements in training data make it an alternative when the annotation budget is limited. The evaluation code and datasets are given for future benchmarking at https://github.com/deepakri201/AbdoBench, along with inference code and weights for ABDSynth.