#$%^&*AU2020101836A420200924.pdf#####ABSTRACT The invention provides a method for generating femoral X-ray films based on deep learning and digital reconstruction of radiological images, which comprises the following steps: performing deep multi-task regression through a three-dimensional convolution neural network model, automatically extracting CT slices containing trochanter and femoral medial and lateral condyles, segmenting images of femoral trochanter medial and lateral condyles by using a conditional generation antagonistic neural network, performing three-dimensional surface reconstruction on the two areas, calculating the vertices of trochanter and femoral medial and lateral condyles by calculating Gaussian curvature, After calculating the angle between the plane of three points and the horizontal plane, the final angle that needs to be rotated and corrected is obtained. By digitally reconstructing the radiological image, the Xray film analog image at the best position is obtained, replacing the film image used by the traditional CT analog positioning machine. According to the problems that the femoral position in the digital reconstruction radiological image can only be manually calibrated by doctors, the intelligent level is not high, the calibration stability is poor, and the actual needs cannot be met, the computer-aided method is used for correcting the femoral CT film and simulating the X-ray film, which can promote the intelligentization of medical equipment.-2/7Image M qsands lice marking of femom1CT samples Constding 3D CNN fir keyslie recognition Segmerdaton oflesse tre andmedial and latemlfemoml based onCGAN Three-dimensioalreconstnrtionoftrcr terandfeamoml e surface Surface Gaussian curvature calculationand vertex solution Three-dimensimlreconstndion ofthe sudsce ofthe less trhter and femml co e DigitalreconstdionofradL imags Gerration of femx x-ray fim Fig. 2
