{
  "id": 19704048,
  "title": "DM4ISL: high-fidelity and uncertainty-aware in silico labeling of fluorescence images via diffusion models",
  "url": "https://urgent.news/2026/11/12/dm4isl-high-fidelity-and-uncertainty-aware-in-silico-labeling-of",
  "topic": "ai",
  "section": "AI",
  "published": "2026-11-12T00:00:00.000Z",
  "source": {
    "name": "bioRxiv",
    "slug": "biorxiv",
    "url": "https://www.biorxiv.org/content/10.64898/2026.11.10.765545v1?rss=1"
  },
  "original_language": "en",
  "account": "In the realm of cell biology, the ability to accurately predict organelle-specific localizations from label-free images could revolutionize longitudinal and multiplexed live-cell imaging. This is where DM4ISL (Diffusion Model for In Silico Labeling) comes into play. This high-fidelity and uncertainty-aware model effectively tackles the limitations of current techniques, such as producing blurry or structurally inaccurate predictions. DM4ISL achieves this by resolving the intrinsic entanglement of biological signal and noise that is inherent to fluorescence imaging.\n\nDM4ISL capitalizes on the iterative nature of diffusion by employing a step-wise self-ensemble averaging approach. This method helps to eliminate stochastic fluctuations, thereby preventing the common problem of overfitting to noise that typically degrades predictions from standard final-step models. In rigorous benchmarking across six organelles, DM4ISL consistently outperforms state-of-the-art architectures in terms of structural fidelity. This advantage has been confirmed through both pixel-based and application-specific metrics.\n\nMoreover, DM4ISL introduces a novel feature that generates localized uncertainty maps by calculating the standard deviation across intermediate denoising steps. This innovative approach automatically detects erroneous predictions at inference time, providing a robust method for trustworthy, data-driven discovery in cell biology. By delivering high-fidelity predictions alongside built-in reliability metrics, DM4ISL sets a new standard for accurate and reliable in silico labeling of fluorescence images.",
  "summary": "In silico labeling predicts organelle-specific localizations from label-free images, potentially enabling longitudinal and multiplexed live-cell imaging. Here we present DM4ISL (Diffusion Model for In Silico Labeling), a high-fidelity and uncertainty-aware model that overcomes current limitations, such as blurry or structurally inaccurate predictions, by resolving the intrinsic entanglement of…",
  "key_points": [],
  "editors_take": null,
  "illustration": null,
  "coverage": {
    "outlets": 1,
    "also_reported_by": []
  },
  "ai_generated": true,
  "disclaimer": "Summaries, key points and the editor’s take are written by software from other outlets’ reporting and may contain errors — always check the linked original."
}