The invention provides a liver CT registration method and system, and the method comprises the steps: obtaining a cross section image and a coronal plane image of liver CT inspection; inputting the cross section image and the coronal plane image into a trained CNN (Convolutional Neural Network), and obtaining the positions of the key points on the cross section image and the coronal plane image; carrying out the extending of the key points, so as to enable a tetrahedron formed by the extended key points to cover the whole liver region; respectively building a liver standard space and a liver scanning space according to a former CT inspection result and a latter CT inspection result, carrying out the triangulation of the liver standard space and the liver scanning space through the extendedkey points, and obtaining a first tetrahedron set and a second tetrahedron set; calculating the corresponding coordinates of a known point in the liver standard space, so as to map the latter CT inspection result to a first CT inspection result and to enable the liver parts in two CT inspection results to be overlapped. The method solves a technical problem that a conventional image recognition device is not accurate in the processing of a CT image.