Scholarlink

<p>This study aims to develop and evaluate an artificial intelligence (AI)-based diagnostic system for the diagnosis of developmental dysplasia of the hip (DDH) in infant hip ultrasonography. The Graf algorithm was employed to develop an automated model for diagnosing DDH, resulting in a DDH-assisted AI model with an average Graf angle error rate of 0.21 compared to expert diagnostics. NASNetMobile achieved the highest Area Under the Curve (AUC) of 0.864 (95% CI, 0.850–0.878), closely followed by MobileNetV1, DenseNet121, EfficientNetV2B0, NASNetMobile, and ResNet50. UnestedUNet demonstrated the highest overall performance, achieving Dice coefficients of 0.794 and a runtime of 40.078 ms, demonstrating its strong segmentation accuracy with moderate computational demands. DeepLabV3Plus, a handheld ultrasound device integrated with a smartphone, demonstrated a robust and efficient segmentation performance. This study highlights the transformative potential of integrating AI into portable ultrasound devices, enabling accurate, efficient, and accessible diagnostic solutions.</p>

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

Automated Neonatal Hip Ultrasound System for Diagnosing Developmental Dysplasia of Hips Using Assistive AI

  • 期刊论文

  • 发表日期：

    2025年4月23日

查看全文
  • Young Seop Lee,
  • Young Jae Kim,
  • Jeong Won Ryu,
  • Su Yeol Lee,
  • Kwang Gi Kim

摘要

This study aims to develop and evaluate an artificial intelligence (AI)-based diagnostic system for the diagnosis of developmental dysplasia of the hip (DDH) in infant hip ultrasonography. The Graf algorithm was employed to develop an automated model for diagnosing DDH, resulting in a DDH-assisted AI model with an average Graf angle error rate of 0.21 compared to expert diagnostics. NASNetMobile achieved the highest Area Under the Curve (AUC) of 0.864 (95% CI, 0.850–0.878), closely followed by MobileNetV1, DenseNet121, EfficientNetV2B0, NASNetMobile, and ResNet50. UnestedUNet demonstrated the highest overall performance, achieving Dice coefficients of 0.794 and a runtime of 40.078 ms, demonstrating its strong segmentation accuracy with moderate computational demands. DeepLabV3Plus, a handheld ultrasound device integrated with a smartphone, demonstrated a robust and efficient segmentation performance. This study highlights the transformative potential of integrating AI into portable ultrasound devices, enabling accurate, efficient, and accessible diagnostic solutions.