{
  "id": 18207194,
  "title": "Cross-cancer knowledge transfer for lymph node metastasis prediction: Applicability of an axillary lymph node-based deep learning model",
  "url": "https://urgent.news/2026/11/04/cross-cancer-knowledge-transfer-for-lymph-node-metastasis-prediction",
  "topic": "health",
  "section": "Health & Medicine",
  "published": "2026-11-04T00:00:00.000Z",
  "source": {
    "name": "bioRxiv",
    "slug": "biorxiv",
    "url": "https://www.biorxiv.org/content/10.64898/2026.11.04.764191v1?rss=1"
  },
  "original_language": "en",
  "account": "The ability of a deep learning model to predict lymph node metastasis across different cancer types has been demonstrated in a recent study. Researchers developed a DenseNet-121 ensemble model using a large dataset of axillary lymph nodes from breast cancer patients. This model was then tested on CT images of lymph nodes from four different anatomical regions—cervical, mediastinal, upper-abdominal, and lower-abdominal—showing above-chance generalizability across all sites.\n\nThe study found that the model's performance varied depending on the specific site of the lymph node being assessed. For cervical lymph nodes, a 40% perinodal margin yielded the highest accuracy (AUROC 0.939, 95% CI 0.875-0.989), while mediastinal lymph nodes performed best with a narrower 10% margin (AUROC 0.796, 95% CI 0.672-0.903). Upper and lower abdominal lymph nodes showed optimal performance at a 30% margin (AUROC 0.814 for upper, AUROC 0.917 for lower).\n\nBy adjusting the perinodal margin, the single ALN-trained DL model was able to accurately predict lymph node metastasis across different cancer sites. This research supports the feasibility of transferring knowledge between different cancers and the potential of using a generalizable backbone for assessing nodal metastasis in a pan-cancer context. The findings have significant clinical implications for improving cancer staging and treatment planning.",
  "summary": "Evaluation of lymph node (LN) metastasis is critical for cancer staging and treatment planning. Since radiologic hallmarks of metastatic LNs are shared across primary cancers, a computed tomography (CT)-based deep learning (DL) classifier developed from one anatomy may transfer to others. This study built a DenseNet-121 ensemble model using clinical-implication-applied preprocessing and evaluated…",
  "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."
}