Automatic Diagnosis System for Lumbar Intervertebral Disc

Aiming at the problem of low accuracy and efficiency of traditional lumbar intervertebral disc diagnosis, an application research on automatic diagnosis of lumbar intervertebral disc based on improved faster R-CNN is proposed. Since the original data set is too small, first use data enhancement and other means to expand the data set; then set up the control group and the experimental group separately. The control group uses the original network parameters, and the experimental group uses the modified network parameters. The backbone networks of the two groups of experiments are VGG-16, ResNet-50 and MobileNet networks in sequence, and they are tested under each network respectively, and the corresponding detection accuracy is obtained. In the experiment process, first input the picture containing the lumbar intervertebral disc, then extract the features after the backbone network convolution, then send it to the lumbar intervertebral disc region proposal network to extract the candidate boxes, and finally enter the region of interest pooling network to complete the detection and classification. Finally, choose the best detection network for clinical diagnosis. The experimental results show that: compared with the original diagnosis method, the improved method has a maximum increase of 12.9% in diagnosis accuracy and a maximum increase of 3.6 times in efficiency.