Automatic segmentation process of a 3D medical image by several neural networks through structured convolution according to the geometry of the 3D medical image

摘要

This invention concerns an automatic segmentation method of features, such as anatomical and pathological structures or instruments, which are visible in a 3D medical image of a subject, composed of voxels.Said method being characterised in that it consistsin providing a global software means or arrangement combining N different convolutional neural networks or CNNs, with N≥2, and having a structured geometry or architecture adapted and comparable to that of the image volume,and in analysing voxels forming said volume of the 3D image according to N different reconstruction axes or planes, each CNN being allocated to the analysis of the voxels belonging to one axis or plane.