Scholarlink

In a time where healthcare is rapidly advancing toward intelligent, automated solutions, the early detection and grading of degenerative diseases like Knee Osteoarthritis (KOA) remain a critical challenge. KOA is a progressive joint disorder that significantly impacts mobility and quality of life. Early and accurate grading of KOA is crucial for timely intervention but manual examination of X-ray images for KOA grading is not only labor-intensive but also prone to diagnostic variability and time-consuming. In light of these obstacles, there exists a pressing necessity to create automated diagnostic measures. This study presents an AI-assisted system that classifies KOA severity directly from knee X-ray images, enabling faster, more consistent, and scalable diagnosis. We propose an Ordinal ensemble deep learning framework that leverages transfer learning and ordinal regression via the CORN (Conditional Ordinal Regression for Neural Networks) strategy. Four pre-trained convolutional neural networks (CNNs): VGG-19, ResNet-34, DenseNet-121, and DenseNet-161 are fine-tuned on a KOA dataset and combined into an ensemble model to output a quantitative Kellgren–Lawrence (KL) grade. Our experimental results suggest an overall testing accuracy near 71.2%. Our results show an improved performance from the existing work in the literature.

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

Ordinal Ensemble Deep Learning for Knee Osteoarthritis Detection Using X-Ray Imaging

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • Deepak T. Mane,
  • Amol Kamble,
  • Deepak More,
  • Yash Deshmukh,
  • Vrushabh Patil,
  • Dewanshu Gakhare

摘要

In a time where healthcare is rapidly advancing toward intelligent, automated solutions, the early detection and grading of degenerative diseases like Knee Osteoarthritis (KOA) remain a critical challenge. KOA is a progressive joint disorder that significantly impacts mobility and quality of life. Early and accurate grading of KOA is crucial for timely intervention but manual examination of X-ray images for KOA grading is not only labor-intensive but also prone to diagnostic variability and time-consuming. In light of these obstacles, there exists a pressing necessity to create automated diagnostic measures. This study presents an AI-assisted system that classifies KOA severity directly from knee X-ray images, enabling faster, more consistent, and scalable diagnosis. We propose an Ordinal ensemble deep learning framework that leverages transfer learning and ordinal regression via the CORN (Conditional Ordinal Regression for Neural Networks) strategy. Four pre-trained convolutional neural networks (CNNs): VGG-19, ResNet-34, DenseNet-121, and DenseNet-161 are fine-tuned on a KOA dataset and combined into an ensemble model to output a quantitative Kellgren–Lawrence (KL) grade. Our experimental results suggest an overall testing accuracy near 71.2%. Our results show an improved performance from the existing work in the literature.