{
  "patent_number": "US 20260080696",
  "country": "US",
  "title": "AI System for Analyzing Medical Images Using Magnification-Aligned Transformers",
  "original_title": "Method and system for analyzing pathological images based on magnification-aligned transformer (mat)",
  "summary": "This 2025 patent describes an AI system that analyzes medical images, like those from pathology slides, by using a special 'Transformer' model that understands how image details change with magnification.",
  "what_it_does": "This patent details a method for analyzing pathological images, such as those from whole-slide images (WSIs) used in medical diagnosis. First, it identifies and isolates the actual tissue regions within these large images, discarding blank or unusable areas. Then, it breaks down these tissue regions into smaller 'patches.' The core of the invention is a 'Magnification-Aligned Transformer' (MAT) classification network model. This model has two main parts: one that aligns features from images taken at different magnifications (like zooming in or out) and another that uses both global (big picture) and local (detailed) attention mechanisms, combining the strengths of Transformers for overall understanding and Convolutional Neural Networks (CNNs) for fine details. This trained model then predicts a classification for the pathological image, helping doctors diagnose conditions. For example, it could help identify cancerous cells in a tissue sample.",
  "what_it_does_not_cover": [
    "Analyzing images that are not pathological or medical in nature.",
    "Methods that do not involve breaking down whole-slide images into patches.",
    "AI models that do not use a Transformer architecture.",
    "Systems that do not specifically align features across different image magnifications.",
    "Analysis that relies solely on global features without considering local details."
  ],
  "filed": "2025-11-25",
  "granted": null,
  "expires": "2045-11-25",
  "status": "active",
  "holder": "GUANGDONG PROVINCIAL PEOPLE'S HOSPITAL",
  "holder_url": "https://patentbrief.org/company/guangdong-provincial-peoples-hospital",
  "inventors": [
    {
      "name": "Chu HAN",
      "url": "https://patentbrief.org/inventor/chu-han"
    },
    {
      "name": "Zaiyi LIU",
      "url": "https://patentbrief.org/inventor/zaiyi-liu"
    },
    {
      "name": "Bingchao ZHAO",
      "url": "https://patentbrief.org/inventor/bingchao-zhao"
    },
    {
      "name": "Jiatai LIN",
      "url": "https://patentbrief.org/inventor/jiatai-lin"
    },
    {
      "name": "Yanqi HUANG",
      "url": "https://patentbrief.org/inventor/yanqi-huang"
    },
    {
      "name": "Zhenwei Shi",
      "url": "https://patentbrief.org/inventor/zhenwei-shi"
    }
  ],
  "times_cited": 0,
  "tags": [
    "biotech",
    "medical_devices",
    "ai_ml",
    "software"
  ],
  "abstract": "A method for analyzing pathological images based on a magnification-aligned transformer (MAT) is provided, in which a pathological image dataset is identified and segmented to obtain pathological image patches; the pathological image patches is screened to obtain a patch set; an MAT classification network model including a self-supervised magnification alignment module and a global-local Transformer classification module is constructed; the MAT classification network model is trained for self-supervised magnification alignment using the patch set in the self-supervised magnification alignment module; the MAT classification network model is further trained using a convolutional neural network (CNN)-transformer; and a pathological image classification prediction result is obtained using the trained MAT classification network model. A system for implementing such method is also provided.",
  "url": "https://patentbrief.org/patent/us/20260080696/method-and-system-for-analyzing-pathological-images-based-on-magnification-align",
  "markdown_url": "https://patentbrief.org/patent/us/20260080696/method-and-system-for-analyzing-pathological-images-based-on-magnification-align/md",
  "google_patents_url": "https://patents.google.com/patent/US20260080696",
  "relatedPatents": [
    {
      "patentNumber": "10452978",
      "countryCode": "US",
      "title": "How AI Models Understand Language Using 'Attention'",
      "url": "https://patentbrief.org/patent/us/10452978/transformer-attention-mechanism"
    },
    {
      "patentNumber": "3789832",
      "countryCode": "US",
      "title": "Using Nuclear Magnetic Resonance to Detect Cancer in Tissue",
      "url": "https://patentbrief.org/patent/us/3789832/mri-magnetic-resonance-imaging"
    },
    {
      "patentNumber": "20220383078",
      "countryCode": "US",
      "title": "Adapting AI Models to Fit Device Resources",
      "url": "https://patentbrief.org/patent/us/20220383078/data-processing-method-and-related-device"
    },
    {
      "patentNumber": "9965705",
      "countryCode": "US",
      "title": "How AI Uses Question-Guided Attention to Answer Questions About Images",
      "url": "https://patentbrief.org/patent/us/9965705/systems-and-methods-for-attention-based-configurable-convolutional-neural-networks-abc-cnn-for-visual-question-answering"
    },
    {
      "patentNumber": "12574477",
      "countryCode": "US",
      "title": "Training AI Models Across Different Computers",
      "url": "https://patentbrief.org/patent/us/12574477/distributed-deep-learning-using-a-distributed-deep-neural-network"
    }
  ]
}