3DSRNet: 3-D Spine Reconstruction Network Using 2-D Orthogonal X-Ray Images Based on Deep Learning

Orthopedic spine disease is one of the most common diseases in the clinic. The diagnosis of spinal orthopedic injury is an important basis for the treatment of spinal orthopedic diseases. Due to the complexity of the spine structure, doctors usually need to rely on orthopedic computed tomography (CT) image data for accurate diagnosis. In some cases, such as poor areas or in emergency situations, it is difficult for doctors to make accurate diagnoses using only 2-D x-ray images due to lack of 3-D imaging equipment or time crunch. Therefore, an approach based on 2-D x-ray images is needed to solve this problem. In this article, a novel 3-D spine reconstruction technique based on 2-D orthogonal x-ray images (3DSRNet) is designed. 3DSRNet uses a generative adversarial network (GAN) architecture and novel modules to make 3-D spine reconstruction more accurate and efficient. Spine reconstruction convolutional neural network (CNN)-transformer framework (SRCT) is employed to effectively integrate local bone surface information and long-range relation spinal structure information. Spine reconstruction texture framework (SRTE) is used to extract spine texture features to enhance the effect of pixel-level reconstruction. Experiments show that 3DSRNet achieves excellent 3-D spine reconstruction results on multiple metrics including peak signal-to-noise ratio (PSNR) (45.4666 dB), structural similarity index (SSIM) (0.8850), cosine similarity (CS) (0.7662), mean absolute error (MAE) (23.6696), mean squared error (MSE) (9016.1044), and learned perceptual image patch similarity (LPIPS) (0.0768).