FIELD: computer engineering.
SUBSTANCE: automatic segmentation of elements, such as anatomical and pathological structures or devices, which are visible on a three-dimensional medical image of an object consisting of voxels. This method consists in providing a common software tool or device that combines N different convolutional neural networks or CNNs in the number of N≥2 networks, and having a structured geometry or architecture adapted to and comparable to that of the image volume, and in the analysis of voxels forming said three-dimensional image volume in N different reconstruction axes or planes, with each CNN dedicated to the analysis of voxels belonging to the same axis or plane.
EFFECT: increasing the accuracy of segmentation of a three-dimensional medical image.
16 cl, 8 dwg
