Detection of Spinal Fracture Lesions Based on Improved Faster-RCNN

Because of the problem that the complexity of spine CT images, the irregular shape of vertebral boundary, low contrast, noise and unevenness in images, meanwhile there are artificial deviations and low efficiencies in clinic, which needs doctors' prior knowledge and clinical experience to determine lesions location in CT images, so it can not meet the clinical realtime needs. In this paper, we use deep learning to process the CT images of spine, and to detect and locate lesion of (cervical fracture, cfracture), (thoracic fracture, tfracture), (lumbar fracture, lfracture) by the improved Faster-RCNN[1]. Through improving the RPN network in Faster-RCNN and changing the number of anchor, we choose appropriate length-width ratio to improve detection efficiency and accuracy. The experiment shows the results are more accurate, and mAP (mean average precision) of detection algorithm is 73.3%, detection rate is 0.03810 seconds per detection, which can basically meet the clinical real-time needs.