Improved deep joint model with tuned improved LeNet for spinal cord segmentation and injury detection using CT images

Spinal cord injury (SCI) is a serious global health issue that often leads to temporary or permanent loss of sensation and motor function. Early and accurate detection is crucial for improving patient outcomes. This study proposes a novel SCI detection model using CT images. The workflow includes preprocessing with Gaussian filtering for noise reduction and Z-score normalization for intensity adjustment. A Modified Deep Joint (MDJ) segmentation approach is then used to accurately segment spinal cord regions, followed by the extraction of connectivity, SLBT, color, statistical, and improved correlation features. For classification, an enhanced LeNet model is optimized using a new WaOI - ZOA algorithm. Experimental results show that the proposed WaOI - ZOA + M-LeNet model achieves a maximum accuracy of 0.9652, outperforming other models such as WaOI - ZOA + Squeeze Net (0.8641), WaOI - ZOA + LeNet (0.8564), and WaOI - ZOA + DCNN (0.7569). These findings confirm the effectiveness of the proposed approach for precise SCI detection.