HYBRID DEEP SPIKING RESIDUAL BELIEF NETWORK FOR INJURY DETECTION AND CLASSIFICATION

The defects in the spinal cord (SC) are caused due to vertebral fracture and this results in sensation and muscle function losses. The nervous system is connected with SC, and the brain. The spinal cord injury (SCI) is caused due to injuries in the spinal tracks and this affects the lifetime of the individuals which leads to lifetime disabilities. The neurological condition of the individuals is enhanced by the early detection and the diagnosis of SCI. However, various research works and clinical assessments were limited by poor SC segmentation process, which is valuable for disease diagnosis. Therefore, this paper introduces a Deep Spiking Residual Belief Network (DSResBNet) for SCI detection and categorization. Initially, input image is preprocessed and given to the SC segmentation phase and the segmentation is done using deep joint segmentation technique. Then, disk localization is performed utilizing active contour model (ACM). Further, feature extraction is conducted for gathering essential features from the input. Finally, SCI prediction and categorization are done with DSResBNet, which is modeled by the incorporation of spiking residual network (S-ResNet) and deep belief network (DBN). The evaluation results reveal that the DSResBNet achieved higher levels of accuracy, sensitivity, and specificity of 90.831%, 91.423%, and 90.816%.