Automatic segmentation algorithm of femur and tibia based on Vnet-C network

Segmentations of the femur and tibia are very important in the development of an orthopedic surgery plan, but the current segmentation method for the femur and tibia requires a large amount of user interaction, which requires too much time and effort for the doctor. In order to make the segmentation of femur and tibia more efficient, we propose an automatic segmentation algorithm based on Vnet-C network. Firstly, the algorithm takes the three-dimensional volume data of the femur and tibia as input, and uses the three-dimensional convolution to extract the three-dimensional features of the femur and tibia; Then, we use residual learning to construct a deeper network to improve the accuracy of automatic segmentation. To verify the algorithm's result, the segmentation experiment was performed on the lower extremity CT scan of 68 patients, and we evaluate the segmentation results through the following metrics: Precision(P), recall rate (R), F1 value, Dice score (DC) and Jaccard index (IOU). The results show that the Vnet-C network can achieve accurate automatic segmentation of the femur and tibia.