The central idea behind CROWN is borrowed from one of the most consequential trends in modern artificial intelligence: the foundation model. Rather than training a bespoke neural network for each individual diagnostic problem, a foundation model is pretrained once on an enormous, unlabeled corpus and then adapted cheaply to many downstream applications. Foundation models have already transformed natural-language processing and, more recently, computational pathology of tissue sections. Cytology, however, has lagged behind. Cytological preparations differ fundamentally from histology: they consist of sparsely distributed, individually preserved cells on cluttered, often stain-inconsistent backgrounds rather than dense tissue architecture. Prior cytology-specific models have tended to be narrow, task-specific classifiers that generalize poorly across staining protocols, scanners and anatomical sites.

To build CROWN, the team assembled a pretraining corpus of more than ten million cytology image patches drawn from multiple institutions and anatomical sites. Crucially, the model was trained with a DINOv2-based self-supervised framework, a strategy that requires no manual annotations whatsoever during pretraining. In self-supervised learning, the network learns by solving pretext problems on the raw images themselves, for example by learning to recognize that differently cropped or augmented views of the same image patch should map to similar internal representations. This forces the model to discover the visual grammar of cytology on its own: nuclear contours, chromatin texture, cytoplasmic staining, cell-to-cell arrangements and the characteristic appearance of malignant, benign and reactive cells, all without a single pathologist-supplied label.

Architecturally, CROWN follows the vision transformer paradigm that has become the standard for large-scale visual representation learning, in which an image is divided into small patches that are processed through attention mechanisms so that every patch can influence every other. The authors pretrained the network at scale and then evaluated its frozen features across an unusually broad benchmark suite: 202 task-setting combinations spanning patch-level classification, cell segmentation, object detection, image retrieval and slide-level prediction. The evaluation cohorts included private institutional collections, among them cohorts containing lymph node metastasis samples and cervical screening specimens, alongside a battery of public datasets covering cervical cytology, thyroid aspirates, effusion cytology and hematology preparations. The benchmark data supporting the analyses have been made publicly available, a transparency measure the researchers argue is essential for credible comparison of future cytology encoders.

The headline results are striking. Across the diverse benchmark settings, CROWN achieved the best overall performance among established pretrained encoders. In 48 patch-level classification evaluations the model exceeded 95 percent accuracy, and in 22 of those evaluations accuracy surpassed 98 percent. Zero-shot evaluations, in which the model classifies images it has never seen labeled examples of, and linear probing evaluations, in which only a simple linear classifier is fitted on top of the frozen features, both demonstrated that the representations learned during unsupervised pretraining capture diagnostically meaningful information without task-specific fine-tuning of the backbone.

The versatility of the learned features extended well beyond classification. On public thyroid and cervical segmentation datasets, CROWN-based models delivered strong DICE scores at both the single-cell and whole-image levels, indicating that the encoder preserves fine-grained spatial information needed to trace individual cell boundaries. On blood-cell object detection benchmarks, CROWN features supported competitive average precision, and on image retrieval tasks spanning more than a dozen anatomical sites and fine-grained subtypes across the two institutional cohorts, the model retrieved diagnostically relevant neighbor images with high mean-average accuracy, a capability with potential value in case-based decision support and digital consultation.

Perhaps most consequential for real-world screening are the slide-level experiments. Whole cytology slides contain thousands of cells, and labeling them individually is prohibitively expensive. CROWN was evaluated in weakly supervised and few-shot settings, in which the model must classify an entire slide from slide-level labels alone or from only a handful of labeled examples. Using prototype-based classification, in which query slides are assigned to the class whose representative feature prototype lies nearest in the learned embedding space, CROWN maintained strong accuracy even with minimal supervision. Grad-CAM visualizations showed that the model concentrates its attention on diagnostically relevant regions, offering pathologists a window into why a prediction was made, an interpretability feature considered essential for clinical acceptance of artificial intelligence in pathology.

The study also compared CROWN directly against publicly released task-specific models on representative public cytology datasets, as well as against a family of established pretrained encoders including general-purpose vision models and histology-oriented foundation models. CROWN outperformed or matched these alternatives across most benchmark combinations, reinforcing the authors’ argument that a single, cytology-specialized visual backbone can displace fragmented collections of narrow models. The benchmark itself, with 202 distinct task-setting combinations evaluated under five-fold cross-validation and reported with standard error measurements, ranks among the most comprehensive evaluations ever assembled for computational cytology and sets a reference point that subsequent work will be measured against.

The code and model weights for CROWN have been released for academic research purposes on GitHub and the Hugging Face model hub, lowering the barrier for laboratories worldwide to adapt the model to local datasets and staining protocols. The researchers are candid about limitations: the pretraining data contain patient-derived material and remain restricted by institutional ethics regulations and privacy requirements, so the pretraining corpus itself is not publicly available. The team also cautions that prospective clinical validation will be needed before CROWN can support real diagnostic workflows. Nevertheless, the message is clear. By demonstrating that self-supervised pretraining at scale can yield a single visual backbone that generalizes across organs, preparations, scanners and task types, CROWN establishes a credible template for universal computational cytology, a field in which the demand for expert eyes far outstrips the supply. If follow-up clinical studies bear out its benchmark performance, models of this kind could help triage slides, flag suspicious cells and extend expert-level screening to regions where cytologists are scarce.

Subject of Research: A self-supervised visual foundation model for universal computational cytopathology

Article Title: A universal visual foundation model for computational cytopathology

Article References: A universal visual foundation model for computational cytopathology. (n.d.). https://doi.org/10.1038/s43018-026-01240-0

Image Credits: AI Generated

DOI: 10.1038/s43018-026-01240-0

Keywords: cytopathology, foundation model, self-supervised learning, DINOv2, vision transformer, cervical cancer screening, digital pathology, deep learning, cancer diagnosis, medical imaging AI, image retrieval, weakly supervised learning

At the core of this research lies the concept of foundation models—large, pre-trained neural networks capable of generalizing to various tasks with minimal additional training. These models have gained traction in numerous fields, such as natural language processing and computer vision; however, their application in computational pathology has been relatively underexplored. The team’s work represents an essential examination into how these models can be harnessed to extract relevant features from medical images, thus aiding pathologists who often operate in environments constrained by time and resources.

The highlighted area of their research is weakly supervised learning, a paradigm particularly suited to medical pathology due to the often limited availability of annotated datasets. In many cases, medical images are abundant, yet labels indicating specific pathologies are scarce due to the labor-intensive process of manual annotation by expert pathologists. The innovative approach described in the study takes advantage of this abundance of unannotated or weakly annotated data. By utilizing foundation models, the researchers demonstrate that it is possible to train robust models effectively without the need for extensive labeled datasets.

Through rigorous benchmarking, the team evaluates several foundation models to determine their effectiveness as feature extractors. The results indicate that these models not only improve the accuracy of diagnostic predictions but also significantly reduce the time required for image analysis. By leveraging unsupervised or weakly supervised data, the researchers show that foundation models can capture intricate patterns and features that traditional diagnostic methods might overlook. The performance improvements realized through this methodology could lead to more timely interventions and better patient outcomes.

Another aspect of this research underscores the interpretability of the foundation models employed. As pathologists increasingly rely on artificial intelligence tools, understanding how these models arrive at specific conclusions becomes crucial. The study addresses this need by incorporating explainability frameworks, providing insights into the decision-making processes of the AI systems. This transparency fosters trust among medical professionals, allowing them to utilize AI tools confidently in clinical settings.

Furthermore, the implications of this research extend to the potential democratization of advanced diagnostic tools. Traditional diagnostic techniques often require significant resources, both in terms of technology and expert personnel. However, by harnessing the power of foundation models, healthcare systems, especially those in resource-limited settings, could gain access to proficient diagnostic tools. This would bridge gaps in healthcare equity, ensuring that high-quality imaging analysis is not a privilege reserved solely for well-funded organizations.

Privacy and ethical considerations remain pivotal in discussions surrounding AI in healthcare. The team’s research addresses these complexities by emphasizing the importance of incorporating ethical guidelines in the deployment of AI tools. By adhering to best practices and regulatory standards, the integration of foundation models into clinical workflows can be navigated responsibly, thereby safeguarding patient data while maximizing the potential benefits of technology.

As the healthcare industry grapples with the dual challenges of increasing patient demands and a shortage of skilled professionals, the findings from Neidlinger and colleagues point to a promising future. The successful application of foundation models within weakly supervised computational pathology not only highlights the capabilities of AI but also reinforces the necessity for continued research and development in this domain. As these technologies advance, they hold the potential to significantly alleviate the burden on healthcare systems while enhancing the accuracy and efficiency of diagnoses.

In summary, the research presented by Neidlinger et al. marks a vital step forward in the integration of artificial intelligence within medical diagnostics. By showcasing the utility of foundation models as feature extractors, the study encourages a shift in perspective regarding how we leverage artificial intelligence in the clinical setting. As the intersection of technology and medicine continues to evolve, this work stands as a testament to the innovative approaches that may soon become central to the practice of pathology, ultimately transforming patient care on a global scale.

The research demonstrates that the journey to incorporating artificial intelligence in medicine is not merely about adopting new tools, but also about fostering a collaborative relationship between humans and machines. This relationship, built on trust and transparency, paves the way for more effective healthcare solutions. As researchers delve deeper into the capabilities of foundation models, they open new avenues for exploration, setting the stage for future innovations that may change the face of disease diagnosis and management.

In a world increasingly reliant on data-driven solutions, the contributions of Neidlinger and his team are poised to influence not just the field of computational pathology but the broader landscape of healthcare. As this field progresses, one can foresee a time when artificial intelligence is seamlessly integrated into everyday medical practices, enhancing the expertise of healthcare professionals and ultimately leading to a healthier global population.

The advancements highlighted in this research will pave the way for further studies, encouraging academics and practitioners alike to investigate the boundaries of artificial intelligence in medicine. As we stand on the brink of this new era, the work of these dedicated researchers offers a glimpse into what is possible when cutting-edge technology meets the field of pathology—a convergence that promises to redefine the very nature of medical diagnostics.

In conclusion, the benchmarking of foundation models as feature extractors for weakly supervised computational pathology is not simply an academic exercise; it represents a critical intersection of technology and healthcare. As institutions worldwide grapple with the challenges of modern medicine, the insights gleaned from this research will surely inform future pathways, guiding the integration of AI while addressing the complexities inherent in medical practice.

Subject of Research: The application of foundation models as feature extractors in weakly supervised computational pathology.

Article Title: Benchmarking foundation models as feature extractors for weakly supervised computational pathology.

Article References:

Neidlinger, P., El Nahhas, O.S.M., Muti, H.S. et al. Benchmarking foundation models as feature extractors for weakly supervised computational pathology.
Nat. Biomed. Eng (2025). https://doi.org/10.1038/s41551-025-01516-3

Image Credits: AI Generated

DOI: 10.1038/s41551-025-01516-3

Keywords: foundation models, computational pathology, weakly supervised learning, artificial intelligence in healthcare, medical diagnostics.