{
  "id": 15639525,
  "title": "Why teaching AI about symmetry could make prenatal diagnosis more reliable",
  "url": "https://urgent.news/2026/10/19/why-teaching-ai-about-symmetry-could-make-prenatal-diagnosis-more",
  "topic": "ai",
  "section": "AI",
  "published": "2026-10-19T18:20:00.000Z",
  "source": {
    "name": "MIT News AI",
    "slug": "mit-news-ai",
    "url": "https://news.mit.edu/2026/teaching-ai-about-symmetry-could-make-prenatal-diagnosis-more-reliable-1019"
  },
  "original_language": "en",
  "account": "The development of human fetuses is a complex process vulnerable to genetic, environmental, and metabolic disruptions, which can result in developmental abnormalities. These abnormalities often begin with the brain, the first organ to develop in the fetus after birth. Prenatal ultrasounds may flag potential brain abnormalities for further magnetic resonance imaging (MRI), but imperfect imaging can lead to missed or incorrect diagnoses. To address this issue, researchers from MIT, Boston Children's Hospital, and Inria have developed a deep learning model called E(3)-Pose, which aids in imaging the fetus's head during motion to obtain high-resolution images for radiologists to examine.\n\nOne challenge with fetal MRI is that the baby's movement causes the slices to be taken from different angles, resulting in gaps in the final scan and unusual slice orientation. These gaps can cause radiologists to miss or inaccurately diagnose fetal brain pathologies. E(3)-Pose addresses this issue by focusing on physics, specifically the left-right symmetry of the fetal head, which is embedded in the model architecture. By incorporating this symmetry, E(3)-Pose can more accurately identify the brain's orientation and capture high-resolution images, even when the head is moving during imaging.\n\nThe research team behind E(3)-Pose includes MIT PhD student Ramya Muthukrishnan, her supervisor Sunlin (1966) and Priscilla Chou Professor of Electrical Engineering and Computer Science Polina Golland, and co-authors Borjan Gagoski, Aryn Lee, Elfar Adalsteinsson, Ellen Grant, and Benjamin Billot. Golland highlights the significance of Muthukrishnan's work in combining geometric concepts with deep learning to improve MRI acquisition and diagnosis. The model has potential applications beyond fetal MRI, including imaging other hard-to-image anatomy and autonomous driving applications. Ultimately, Muthukrishnan aims to enhance E(3)-Pose to accommodate imaging of fetuses in the second trimester or younger, as early detection of brain abnormalities can lead to easier intervention.",
  "summary": "A model developed by researchers at MIT and Boston Children's Hospital adjusts scans on the fly to track fetal movement, potentially sparing families repeat hospital visits.",
  "key_points": [
    "Researchers develop E(3)-Pose deep learning model to aid fetal MRI imaging.",
    "Model incorporates left-right symmetry of fetal head to improve image accuracy.",
    "Potential applications extend to other hard-to-imagine anatomy and autonomous driving."
  ],
  "editors_take": null,
  "illustration": null,
  "coverage": {
    "outlets": 2,
    "also_reported_by": [
      {
        "outlet": "MIT News Research",
        "title": "Why teaching AI about symmetry could make prenatal diagnosis more reliable",
        "url": "https://urgent.news/2026/10/19/why-teaching-ai-about-symmetry-could-make-prenatal-diagnosis-more-15641089",
        "published": "2026-10-19T18:20:00.000Z"
      }
    ]
  },
  "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."
}