Scholarlink

<p>Cervical spondylitis refers to the cervical vertebral fractures and degeneration of the cervical spine leading to symptoms like neck soreness and stiffness. Detecting and diagnosing cervical vertebral fractures promptly is crucial, especially in the elderly population where cervical vertebral fractures may be challenging to identify due to degenerative illness and osteoporosis. Traditional imaging methods have limitations in spotting hairline cracks, but recent advancements in deep learning models offer promising solutions. In this work, Convolutional Neural Networks (CNN), Fast R-CNN and transfer learning models, such as Alex-Net and EfficientB6, to predict, classify, and pinpoint the exact location of vertebral fractures were used. The use of computed tomography (CT) has significantly improved adult spine fracture imaging over conventional radiography (x-rays). Our results illustrate the power of the proposed models. CNN had an accuracy of 90.2%, Alex-Net had an accuracy of 93.7% and Efficient Net B6 outperformed all the models with an accuracy of 97.34%. The classification performance was assessed by accuracy, sensitivity, specificity, precision, recall, F1-score, confusion matrix analysis, and AUC-ROC attributes. For better localization of fractures, YOLOV5 also demonstrated viable results with an accuracy of 93.84%. These results indicate the feasibility of developed deep models for cervical vertebral fracture detection and localization in CT images. EfficientB6 and YOLOV5 specifically, appear as being the best candidates to accurately designate the exact fracture point, an indication that these findings could pave the way end or high level clinical management in a timely manner.</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

Deep learning-based detection and localization of cervical vertebral fractures from CT images

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年7月22日

查看全文
  • Lakshmana Rao Kalabarige,
  • Upendra Kumar Potnuru,
  • Srinivasa Kishore Teegala,
  • Venkataramana Guntreddi

摘要

Cervical spondylitis refers to the cervical vertebral fractures and degeneration of the cervical spine leading to symptoms like neck soreness and stiffness. Detecting and diagnosing cervical vertebral fractures promptly is crucial, especially in the elderly population where cervical vertebral fractures may be challenging to identify due to degenerative illness and osteoporosis. Traditional imaging methods have limitations in spotting hairline cracks, but recent advancements in deep learning models offer promising solutions. In this work, Convolutional Neural Networks (CNN), Fast R-CNN and transfer learning models, such as Alex-Net and EfficientB6, to predict, classify, and pinpoint the exact location of vertebral fractures were used. The use of computed tomography (CT) has significantly improved adult spine fracture imaging over conventional radiography (x-rays). Our results illustrate the power of the proposed models. CNN had an accuracy of 90.2%, Alex-Net had an accuracy of 93.7% and Efficient Net B6 outperformed all the models with an accuracy of 97.34%. The classification performance was assessed by accuracy, sensitivity, specificity, precision, recall, F1-score, confusion matrix analysis, and AUC-ROC attributes. For better localization of fractures, YOLOV5 also demonstrated viable results with an accuracy of 93.84%. These results indicate the feasibility of developed deep models for cervical vertebral fracture detection and localization in CT images. EfficientB6 and YOLOV5 specifically, appear as being the best candidates to accurately designate the exact fracture point, an indication that these findings could pave the way end or high level clinical management in a timely manner.