{
  "id": 8937750,
  "title": "Beyond raw brainpower: New study reveals best ways to train AI for clinical care",
  "url": "https://urgent.news/2026/09/21/beyond-raw-brainpower-new-study-reveals-best-ways-to-train-ai-for",
  "topic": "ai",
  "section": "AI",
  "published": "2026-09-21T14:20:01.000Z",
  "source": {
    "name": "Medical Xpress",
    "slug": "medical-xpress",
    "url": "https://medicalxpress.com/news/2026-09-raw-brainpower-reveals-ways-ai.html"
  },
  "original_language": "en",
  "account": "A new study published in the Journal of Medical Internet Research reveals that training artificial intelligence (AI) for use in clinical care requires more than simply leveraging raw technology. To ensure AI can safely and effectively assist with medical diagnoses, triage, and treatment planning, experts believe that adapting existing language models is crucial.\n\nA team of researchers led by Anshum Patel, MD, and Joseph Y Cheung, MD, MS, examined 35 recent studies to understand how different customization methods impact AI performance. Their findings indicate that while standard large language models (LLMs) are capable, they need to be tailored specifically for medical environments to be reliable. When AI models are directly connected to trusted medical databases or retrained on specific clinical guidelines, their accuracy significantly improves, with some systems matching the diagnostic performance of human doctors.\n\nThe most effective approach depends on the specific medical task at hand. For narrow, focused tasks like identifying cancer in medical images, retraining the AI on specific data yields the best results. However, for tasks requiring reasoning through complex guidelines, linking the AI to live databases proves highly effective. The researchers discovered that the best performance emerges from hybrid systems that combine both methods, enabling AI to manage intricate workflows such as stroke triage and oncology cases.\n\nWhile these findings are promising, the researchers emphasize that most studies on AI systems in healthcare are based on historical medical records rather than real-world patient testing. Before widely adopting these advanced tools in hospitals, additional prospective, real-world testing is necessary to ensure patient safety and guarantee that the technology functions reliably across various clinical settings.",
  "summary": "Harnessing generative AI to help doctors make critical medical decisions requires more than raw technology. A new review study in the Journal of Medical Internet Research shows that adapting existing language models is key to ensuring they can safely and effectively assist with clinical diagnoses, patient triage and treatment planning in real-world health care settings.",
  "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."
}