Scholarlink

Down syndrome, is a genetic condition characterized by developmental delays, intellectual disabilities, and recognizable physical features. Early detection during pregnancy is crucial, as it allows parents to make informed decisions and prepare for necessary medical care. The paper offers a non-invasive and user-friendly system to support early screening by analyzing ultrasound images. The proposed system is mainly based on utilizing deep-learning architecture. It consists of two main phases; front-end and back-end. The front end of the proposed system offers a user-friendly interface that guides users through uploading fetal ultrasound images directly from their device. Once an image is submitted, the system displays real-time diagnostic results, indicating whether the case is Standard or Non-Standard. The backend serves as the system’s core processing the uploaded ultrasound images through a preprocessing pipeline, and passing them to a deep learning model built on for prediction. The backend flow begins with the collection of the raw ultrasound dataset, which is then subjected to a preprocessing and undergo noise reduction to enhance clarity and remove artifacts. The resulting preprocessed dataset is then fed into EfficientNetB0 which analyzes each image and outputs a prediction classifying it as either Standard (normal) or Non-Standard (at risk). It also supports ongoing model retraining and performance enhancement as more data becomes available, ensuring sustained reliability and diagnostic accuracy over time. The system demonstrated strong performance, achieving 93% accuracy in detecting Down syndrome from ultrasound images, indicating its potential value in prenatal care.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

A Deep Learning-Based System for Automated Detection and Support of Down Syndrome via Ultrasound Imaging

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • Mark Emad,
  • Sama Mahmoud,
  • Farah Waleed,
  • Omar Ahmed,
  • Mahmoud Yakout,
  • Esmaeil Ahmed,
  • Micheal Zakaria,
  • Amira ElGouhary

摘要

Down syndrome, is a genetic condition characterized by developmental delays, intellectual disabilities, and recognizable physical features. Early detection during pregnancy is crucial, as it allows parents to make informed decisions and prepare for necessary medical care. The paper offers a non-invasive and user-friendly system to support early screening by analyzing ultrasound images. The proposed system is mainly based on utilizing deep-learning architecture. It consists of two main phases; front-end and back-end. The front end of the proposed system offers a user-friendly interface that guides users through uploading fetal ultrasound images directly from their device. Once an image is submitted, the system displays real-time diagnostic results, indicating whether the case is Standard or Non-Standard. The backend serves as the system’s core processing the uploaded ultrasound images through a preprocessing pipeline, and passing them to a deep learning model built on for prediction. The backend flow begins with the collection of the raw ultrasound dataset, which is then subjected to a preprocessing and undergo noise reduction to enhance clarity and remove artifacts. The resulting preprocessed dataset is then fed into EfficientNetB0 which analyzes each image and outputs a prediction classifying it as either Standard (normal) or Non-Standard (at risk). It also supports ongoing model retraining and performance enhancement as more data becomes available, ensuring sustained reliability and diagnostic accuracy over time. The system demonstrated strong performance, achieving 93% accuracy in detecting Down syndrome from ultrasound images, indicating its potential value in prenatal care.