Spinal Fracture Detection Based on Deformable Cascade Network

In order to solve the problems of long CT reading cycles and manual misdiagnosis of spinal burst fracture, the paper proposes an algorithm Deformable Cascade Network(DCNet) for detecting spinal burst fracture grade. Based on Faster RCNN network structure, DCNet algorithm replaces the original VGG 16 by resnet50 with deformable residual module, it adds cascade detector module in the network detection part, thus enhancing the feature extraction ability and detection ability. Through experimental comparison tests, the optimized mean average precision (MAP) of DCNet is 92.8%, precision is 91.7%, recall is 95.7%, and the average detection speed can reach 47.6FPS. The DCNet algorithm proposed in this paper improves the efficiency and accuracy of CT diagnosis of spinal burst fracture, and provides an effective new method for computer-aided diagnosis of spinal burst fracture.