The Flow has not summarised this story yet — read it at arXiv Computer Vision.
arXiv:2608. 03990v1 Announce Type: new Abstract: Synthetic histopathology image generation has emerged as an approach that may address data scarcity in computational pathology, yet current evaluation methodologies may not fully assess synthetic data quality for medical applications.
The paper introduces "Destroy Me", a hybrid framework that generates realistic histopathology artifacts using Stable Diffusion and physics‑based modeling to create six common artifact types. Artifact realism is evaluated with KID and color Wasserstein metrics, and models trained on these augmented images outperform baselines on lung adenocarcinoma classification, achieving a 10.5% relative boost in macro F1‑score and a 15% increase in Cohen’s Kappa. The study highlights that selective, impact‑weighted augmentation is essential for enhancing robustness while preserving subtle diagnostic features.
The paper presents a systematic evaluation of six unsupervised image‑to‑image generative models for virtual Sirius Red staining of mouse liver tissue from routine H&E slides. It benchmarks these models across 54 scaling configurations on a newly released paired H&E‑to‑SR dataset, assessing perceptual, distributional, and task‑specific performance, and further trains the best models into deep ensembles to quantify epistemic uncertainty. The study finds that GAN‑based and diffusion‑based methods differ markedly across these metrics, indicating that reliable virtual staining requires reporting and selecting on all three axes simultaneously.
arXiv:2601. 08127v2 Announce Type: replace-cross Abstract: Expert-annotated training data remains the critical bottleneck for AI in histopathology, particularly for rare pathologies where even dozens of cases may be unavailable.
arXiv:2609.00984v1 Announce Type: cross Abstract: Virtual staining aims to computationally generate target-stained histopathological images while reducing the cost and time associated with convention...
arXiv:2602. 02124v2 Announce Type: replace-cross Abstract: Drug-induced toxicity is a leading cause of preclinical and early-clinical failure, making early detection critical.