Scholarlink

<p>This paper proposes a novel deep learning-based approach for detecting pediatric wrist fractures in radiographs. Our method integrates AC-BiFPN for efficient multi-scale feature fusion and SimAM to emphasize clinically relevant image features, enhancing real-time object detection using YOLOv10. Additionally, we employ the WIoU loss function to improve the model’s generalization capability by minimizing both false positives and, more critically, false negatives. The proposed model was evaluated on the GRAZPEDWRI-DX dataset, comprising 20,327 annotated pediatric wrist radiographs. Our approach achieved significant performance improvements, with a precision of 97.4%, recall of 95.5%, and <InlineEquation ID="IEq1"> <InlineMediaObject> <ImageObject Color="BlackWhite" FileRef="13721_2025_518_Article_IEq1.gif" Format="GIF" Height="16" Rendition="HTML" Resolution="72" Type="Linedraw" Width="50" /> </InlineMediaObject> <EquationSource Format="TEX">\(\text {mAP}_{50}\)</EquationSource> <EquationSource Format="MATHML"><math> <msub> <mtext>mAP</mtext> <mn>50</mn> </msub> </math></EquationSource> </InlineEquation> of 88.5%. Notably, the model demonstrated a strong ability to detect subtle and complex fractures, which are often missed by conventional diagnostic methods. Furthermore, the system exhibited robustness across diverse clinical scenarios while maintaining computational efficiency, making it suitable for real-time deployment in emergency departments. These results suggest that our model not only surpasses traditional fracture detection techniques but also provides a reliable and efficient tool to assist radiologists in pediatric emergency care.</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

Enhanced diagnosis of pediatric wrist fractures using deep learning

  • 期刊论文

  • 发表日期：

    2025年4月23日

查看全文
  • Riadh Bouslimi,
  • Houda Trabelssi,
  • Wahiba Ben Abdessalem Karaa

摘要

This paper proposes a novel deep learning-based approach for detecting pediatric wrist fractures in radiographs. Our method integrates AC-BiFPN for efficient multi-scale feature fusion and SimAM to emphasize clinically relevant image features, enhancing real-time object detection using YOLOv10. Additionally, we employ the WIoU loss function to improve the model’s generalization capability by minimizing both false positives and, more critically, false negatives. The proposed model was evaluated on the GRAZPEDWRI-DX dataset, comprising 20,327 annotated pediatric wrist radiographs. Our approach achieved significant performance improvements, with a precision of 97.4%, recall of 95.5%, and \(\text {mAP}_{50}\) mAP 50 of 88.5%. Notably, the model demonstrated a strong ability to detect subtle and complex fractures, which are often missed by conventional diagnostic methods. Furthermore, the system exhibited robustness across diverse clinical scenarios while maintaining computational efficiency, making it suitable for real-time deployment in emergency departments. These results suggest that our model not only surpasses traditional fracture detection techniques but also provides a reliable and efficient tool to assist radiologists in pediatric emergency care.