Abstract: Cervical spine fractures represent a critical issue in healthcare, especially among older adults, where identifying these injuries can be complicated by existing degenerative conditions. In light of this, a recent study proposes an innovative method that harnesses deep neural networks (DNNs), particularly the U-Net architecture, to streamline the detection of fractures in computed tomography (CT) images. The focus of this methodology is to accurately identify and localize the cervical vertebrae, which is essential for a reliable assessment of fractures. By utilizing the U-Net's strengths in semantic segmentation, the model can effectively outline the boundaries of cervical vertebrae, capturing detailed features and spatial relations within the images.
Additionally, the framework enhances the U-Net's fracture detection capabilities by adding multi-class classification layers, allowing it to differentiate between fractured and intact areas in the segmented vertebrae. This advancement significantly improves the diagnostic accuracy of the approach. Trained on a wide-ranging dataset of cervical spine injuries, this methodology presents substantial clinical benefits, such as facilitating real-time fracture assessments that lead to quicker diagnoses and timely interventions, ultimately enhancing patient outcomes. By tapping into the potential of deep learning, this approach promises to boost both the efficiency and accuracy of cervical spine fracture detection, contributing positively to patient care and treatment results.

Keywords: Key-words: U-Net architecture, computed tomography (CT) images, cervical vertebrae, semantic segmentation.

| DOI: 10.17148/IARJSET.2024.111028