Scholarlink

<p>Accurate detection of bone fractures is essential for effective diagnosis and treatment. Existing deep learning models struggle to predict it correctly due to low image contrast, complex fracture patterns, noise, class imbalance, and optimization challenges. To address these issues, we propose a novel framework combining Adaptive Gamma Equalization for contrast enhancement, Enhanced Gradient Edge Detection for precise fracture boundary delineation (EGED) and a Crack- Region-based Convolutional Neural Network (Crack-RCNN) model with radian activation optimized using Cosine Stochastic Gradient Descent (CosineSGD). This integrated approach improves fracture recognition and localization with an average accuracy of 99% and a minimal loss of 0.1 without overfitting, under fitting and local minima. The framework gives a robust, reliable performance across diverse fracture types.</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

An edge-enhanced neural network for reliable fracture identification

  • 期刊论文

  • 发表日期：

    2025年12月8日

查看全文
  • V. G. Karthiga,
  • P. Dhivya

摘要

Accurate detection of bone fractures is essential for effective diagnosis and treatment. Existing deep learning models struggle to predict it correctly due to low image contrast, complex fracture patterns, noise, class imbalance, and optimization challenges. To address these issues, we propose a novel framework combining Adaptive Gamma Equalization for contrast enhancement, Enhanced Gradient Edge Detection for precise fracture boundary delineation (EGED) and a Crack- Region-based Convolutional Neural Network (Crack-RCNN) model with radian activation optimized using Cosine Stochastic Gradient Descent (CosineSGD). This integrated approach improves fracture recognition and localization with an average accuracy of 99% and a minimal loss of 0.1 without overfitting, under fitting and local minima. The framework gives a robust, reliable performance across diverse fracture types.