Optimization method and system for personalized contrast test based on deep learning

1 . An optimization method of a personalized contrast scan based on deep learning, the method comprising:

obtaining drug information of a contrast medium and body information of a patient, in a contrast enhanced computed tomography (CT) scan;

generating injection information of the drug to be injected into the patient by a predefined algorithm based on the drug information and the body information;

injecting the drug into the patient based on the injection information, and acquiring a medical image by scanning the patient; and

amplifying a contrast component in the medical image by inputting the medical image to a deep learning model trained in advance,

wherein the injection information of the drug, generated by the algorithm, comprises a minimum injection amount that is calculated by: 1) predicting an image contrast enhancement output of the deep learning model prior to acquiring the medical image, and 2) optimizing the injection amount to be sufficient for the deep learning model to generate an image from which a diagnosis can be made after contrast amplification,

wherein obtaining the body information comprises:

acquiring a three-dimensional (3D) image of the patient; and

obtaining, from the 3D image, volume information about a body region excluding a fat region.

2 . The method of claim 1 , wherein the obtaining of the drug information comprises:

recognizing a type of drug based on text information or image information input together with the drug; and

obtaining the drug information corresponding to the drug from a database.

3 . The method of claim 2 , wherein the image information comprises a barcode or a quick response (QR) code.

4 . The method of claim 1 , wherein the acquiring of the 3D image of the patient comprises generating the 3D image based on a combination of two-dimensional (2D) images of the patient.

5 . The method of claim 4 , wherein the generation of the 3D image comprises:

combining a 2D image acquired in up and down directions of the patient and a 2D image acquired in left and right directions of the patient by two-dimensionally scanning the patient; and

generating the 3D image based on the combined images.

6 . The method of claim 1 , wherein the injecting of the drug comprises injecting the drug based on the injection information.

7 . The method of claim 6 , wherein the injecting of the drug comprises automatically adjusting one or both of an injection amount and an injection speed of the drug.

8 . The method of claim 6 , wherein the injecting of the drug comprises outputting and providing one or both of an injection amount and an injection speed of the drug to a user.

9 . The method of claim 1 , wherein the amplifying of the contrast component comprises:

extracting at least one component image of contrast enhanced and unenhanced component images for the medical image by inputting the medical image of the patient to the deep learning model; and

outputting a contrast amplified image for the medical image based on the medical image and the at least one extracted component image.

10 . The method of claim 9 , further comprising: before extracting the component image,

extracting scan information from the medical image; and

selecting at least one deep learning model corresponding to the scan information among a plurality of deep learning models trained in advance.

11 . The method of claim 10 , wherein the extracting of the component image comprises extracting at least one component image for the medical image by inputting the medical image to at least one selected deep learning model.

12 . The method of claim 10 , further comprising: before extracting the scan information,

producing at least one composite component image based on a pair of images from a first training image set;

generating a second training image set based on the at least one composite component image and the first training image set;

extracting the scan information from the second training image set, and grouping the second training image set into a plurality of groups based on a preset rule; and

generating and training a plurality of second deep learning models to respectively correspond to the groups of the grouped second training image set.

13 . The method of claim 12 , wherein the deep learning model selected in the selecting comprises a plurality of deep learning models.

14 . The method of claim 9 , wherein the outputting of the contrast amplified image comprises multiplying the medical image and the at least one component image by preset ratios, respectively, and summing the multiplied images.

15 . The method of claim 9 , wherein the outputting of the contrast amplified image comprises:

multiplying the medical image and the at least one component image by preset ratios, respectively, and summing the multiplied images to generate first and second images; and

outputting a composite color image by applying a preset color tone table to the first and second images.

16 . The method of claim 9 , further comprising: before the extracting of the component image,

producing at least one composite component image based on a pair of images from a first training image set;

generating a second training image set based on the at least one composite component image and the first training image set; and

generating and training a second deep learning model to extract at least one component image by using the second training image set.