The method of estimating the ASPECT score according to the present invention includes a preprocessing step of normalizing and standardizing features of an image dataset by a preprocessor; A segmentation step of the image processing unit separating each lesion within the CT image classified into a supra ganglion level and a ganglion level; And a determining step of determining whether the lesion has a stroke by independently building a neural network for learning a positive/negative image for each lesion by the determination unit; including, wherein the pre-processing step includes convolution of Gaussian blur on the entire patient's brain CT image A noise removal step of removing noise included in the image; A search step of searching for a skull ellipse to find a skull in the brain CT image; An alignment step of aligning the position of the image based on the searched center point of the skull and rotating the image to uniformly align the position and rotation angle of the dataset; And performing a horizontal transformation according to the lesion location (Lesion-Side); wherein the searching step includes searching for an adaptive threshold value in consideration of the distribution of pixel values of the segmented image, and An automatic thresholding step of inducing only the pixel information corresponding to to remain; A contour search step of detecting an edge based on the image for which the thresholding has been completed; And acquiring skull ellipse information for acquiring information on the inner and outer edges of the skull by searching for the skull ellipse from the image remaining only the edge information.
展开▼
