Vertebra Segmentation Based Vertebral Compression Fracture Determination from Reconstructed Spine X-Ray Images

Abstract

The vertebral compression fracture represents the vertebral body deformity appeared over lateral spine imageries. In order to evaluate the vertebral compression fracture (VCF), the vertebral compression ratio (VCR) has to be accurately measured. In most of the existing vertebral segmentation approaches, degraded accuracy, increased possibilities of error and time complexity are found to be the major drawbacks. Hence to conquer these issues and to enhance the overall segmentation performance, rapid automated vertebral segmentation approach is proposed for evaluating the VCR. Initially the reconstructed spine X-ray images are collected and directed over the Hybrid UDA Net architecture from this model, the features are extracted using encoder section of U-net architecture through the adoption of channel attention layer (CaL) and hybrid attention dilated Quantum convolutional layer (HaDQcL). The segmental outcomes are accomplished through the decoder section of U-Net. Based on the extracted features given as the input, exact segmentation of spinal images is attained using Twin attention mechanism called Gated-decoder attention module (GDAM). Through GDAM, the segmented spine X-ray images are obtained with effective results through the fusion of spatial and channel features in decoder attention module. The losses in the neural network are optimized using Amended pelican optimization algorithm (APoA). The diverse stages of VCF are finally analysed through VCR evaluation. The overall accuracy of 98.41%, F1 score of 96.75% and specificity of 99% is obtained by the proposed model whereas the performance is analysed using PYTHON. On comparison of proposed and existing models, the proposed model through segmentation and VCF diagnosis are highly superior.