shepherd siameseconvolutional network for spinal cord injury level classification

Spinal Cord Injury (SCI) has a significant impact on neurological function and quality of life, highlighting the critical importance of precise diagnosis for effective treatment. Current methods for assessing SCI severity frequently depend on the subjective analysis of complex imaging data, which can lead to variability and potential inaccuracies. To address these challenges, a novel deep learning framework called the Shepherd Siamese Convolutional Network (ShSCN-Net) has been proposed for the automated classification of SCI levels. ShSCN-Net combines the capabilities of the Siamese Convolutional Neural Network (SCNN) and Shepherd Convolutional Neural Network (ShCNN) to improve feature extraction and classification. The framework begins by segmenting the spinal cord using a Multi-scale Attention Network (MANet) and precisely locating discs using an active contour model applied to CT images. It then extracts comprehensive features such as connectivity metrics, geometric properties, and Local Ternary Pattern (LTP) descriptors to characterize SCI lesions. Following this, ShSCN-Net accurately detects the level of injury, categorizing SCI into cervical, thoracic, lumbar, and sacral regions. Implemented in Python, ShSCN- Net undergoes rigorous evaluation on the Spineweb Dataset using matrices namely accuracy, True Positive Rate (TPR), as well as True Negative Rate (TNR). Comparative analysis against existing methods shows that ShSCN-Net outperforms in accurately diagnosing SCI levels, indicating its potential to enhance clinical decision-making and advance patient care in the management of spinal injuries. Keywords: SCI, Neurological function, Diagnosis, Imaging data, Deep learning framework, ShSCN-Net, SCNN, ShCNN, MANet, Active contour model, CT images, Connectivity metrics, Geometric properties, LTP descriptors, Spineweb Dataset, TPR, TNR, Clinical decision-making.