Scholarlink

<p>To improve the accuracy of pediatric wrist fracture detection and reduce false positives and missed detections, this study proposes AFE-YOLO, an adaptive feature enhancement algorithm based on YOLOv8. First, receptive field attention convolution was incorporated into the backbone network to enhance feature perception, enabling the model to capture multi-scale fracture features more effectively and optimize computational efficiency. Second, an adaptive hybrid feature enhancement module was introduced in the neck network to improve the recognition of fine-grained lesion regions through feature fusion and semantic representation. Finally, a weighted loss fusion strategy was proposed, combining binary cross entropy and focal loss to address class imbalance and improve sensitivity to hard-to-classify samples. Experimental results demonstrate that AFE-YOLO improves average precision and recall by 4.2 and 7.3 percentage points, respectively, compared with the baseline model on the GRAZPEDWRI-DX dataset, while also reducing computational overhead.</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

AFE-YOLO: An Image Detection Algorithm for Pediatric Wrist Fractures

  • 期刊论文

  • 发表日期：

    2025年12月13日

查看全文
  • Yaolian Song,
  • Tao Huang,
  • Jingmin Tang,
  • Xuanzhi Zhao,
  • Guicai Yu

摘要

To improve the accuracy of pediatric wrist fracture detection and reduce false positives and missed detections, this study proposes AFE-YOLO, an adaptive feature enhancement algorithm based on YOLOv8. First, receptive field attention convolution was incorporated into the backbone network to enhance feature perception, enabling the model to capture multi-scale fracture features more effectively and optimize computational efficiency. Second, an adaptive hybrid feature enhancement module was introduced in the neck network to improve the recognition of fine-grained lesion regions through feature fusion and semantic representation. Finally, a weighted loss fusion strategy was proposed, combining binary cross entropy and focal loss to address class imbalance and improve sensitivity to hard-to-classify samples. Experimental results demonstrate that AFE-YOLO improves average precision and recall by 4.2 and 7.3 percentage points, respectively, compared with the baseline model on the GRAZPEDWRI-DX dataset, while also reducing computational overhead.