Classification of cervical spine disease using convolutional neural network

Cervical spondylosis myelopathy (CSM) is a degenerative disorder of the cervical spine’s disks and joints, leading to neurological disability. Typically, this condition causes increasing discomfort and neurologic deterioration. Cervical spine pain disease classification using deep learning can assist medical professionals in making early and more precise medical diagnosis of cervical abnormalities. The main focus of the assessment for doctors is the classification of cervical spine diseases from X-rays images, which is valuable for patients to indicate the various symptoms they are faced with. Identifying cervical spine pain injuries (CPI) is a critical aspect of the diseases segmentation and classification. Manually explaining the high-dimensional feature space makes it challenging to identify the exact category and level of severity. This paper introduces a Convolutional Neural Networks (CNN) concatenation feature extraction method to help in the diagnosis of cervical spine pain disease from X-ray images. An innovative CNN-concatenation model-based fully connected classifier and a four-class system are applied to classify X-ray image data in this research. The CNN approach is employed to automatically classify X-ray images into healthy (normal) and with cervical spine disease (abnormal). CSM is currently diagnosed mostly through clinical evaluation using imaging methods such as X-ray and computed tomography (CT). Due to its inexpensive cost and minimal radiation dose, X-ray is commonly used. The proposed approach achieved a training accuracy of 99.98%, a validation accuracy of 98.29%, and a testing accuracy of 97.82% of classification from experiments. These results indicate that the image preprocessing, data augmentation, and CNN approaches provide an efficient classification method for identifying and accurately classifying cervical spine diseases. © The Author(s), under exclusive licence to Springer Science+Business Media, LLC, part of Springer Nature 2024.