CPC G06T 7/337 (2017.01) [A61B 6/5205 (2013.01); G06T 3/18 (2024.01); G06T 3/40 (2013.01); G06T 5/10 (2013.01); G06T 7/0014 (2013.01); G06T 7/11 (2017.01); G16H 30/40 (2018.01); G06T 2207/10081 (2013.01); G06T 2207/20132 (2013.01); G06T 2207/30008 (2013.01); G06T 2207/30016 (2013.01); G06T 2207/30101 (2013.01)] | 11 Claims |
1. An information providing method comprising:
a first step of receiving a CT image by a head segmentation unit;
a second step of predicting, by the head segmentation unit, a bone mask of a skull, a brain mask of a brain, and a brain-related vessel mask by passing the CT image through a head segmentation network capable of separating the skull and the brain;
a third step of extracting, by the head segmentation unit, a brain image including only a brain region from the CT image by combining the predicted brain mask and vessel mask;
a fourth step of receiving, by a rigid registration unit, standardized template data from an atlas unit and receiving the brain image from the head segmentation unit by fusing characteristics of specific samples to be used as a reference for aligning a plurality of data into a common space;
a fifth step of performing, by the rigid registration unit, rigid registration and affine registration between the template data and the brain image to acquire a transformation parameter;
a sixth step of applying, by the rigid registration unit, rigid warping to the brain image based on the transformation parameter to acquire a first aligned brain image aligned by a linear transformation;
a seventh step of extracting, by a non-rigid registration unit, a deformation field which is a translation vector for each 3D voxel by inputting a 3D volume of the template data and a 3D volume of the first aligned brain image to a pre-learned non-rigid registration network;
an eighth step of acquiring, by the non-rigid registration unit, a second aligned brain image more accurately registered to the actual brain than a result of the sixth step by performing non-rigid warping that applies grid interpolation in units of voxels to the first aligned brain image based on the deformation field;
a ninth step of extracting, by a neurovascular segmentation unit, a cerebrovascular map from the second aligned brain image by passing the second aligned brain image through a neurovascular anatomy segmentation network; and
a tenth step of providing, by a determination unit, brain-related information based on the cerebrovascular map,
wherein the second step comprises
a 2-1 step of performing pre-processing which is at least one of a resizing operation, a HU windowing operation, and a normalizing operation with respect to the CT image;
a 2-2 step of passing the pre-processed CT image through the head segmentation network;
a 2-3 step of performing post-processing, which is at least one of a brain VOI crop operation and an outlier cluster removing operation with respect to the CT image passing through the head segmentation network; and
a 2-4 step of predicting the bone mask, the brain mask, and the vessel mask based on the post-processed CT image.
|