Coronary artery stent prediction method, device, and recording medium using deep learning based on ultrasound image

A computer-implemented deep learning-based stent prediction method of predicting progress of a stent procedure, the deep learning-based stent prediction method comprising:

• setting, by a stent prediction device, as a region of interest, a region in which a procedure is to be performed among blood vessel regions of a target patient, and obtaining a first intravascular ultrasound (IVUS) image, which is a preprocedural IVUS image of the region of interest;
• obtaining, by the stent prediction device, a plurality of first IVUS cross-sectional images into which the first IVUS image is divided at predetermined intervals;
• extracting, by the stent prediction device, feature information about procedure information of the target patient;
• obtaining, by the stent prediction device, mask image information in which a blood vessel boundary and an inner wall boundary are distinguished from each other, with respect to the plurality of first IVUS cross-sectional images;
• and
• predicting, by the stent prediction device, progress of a stent procedure comprising a postprocedural area of a stent for the target patient, by inputting, into a deep learning model, the plurality of first IVUS cross-sectional images, the feature information, and the mask image information,
• wherein the deep learning-based stent prediction method further comprises training the deep learning model by using, as training data, first processed IVUS images of a plurality of previous patients, wherein the first processed IVUS images are obtained by:
• registering second reference IVUS images, which are postprocedural IVUS images of the plurality of previous patients, with respect to first reference IVUS images, which are preprocedural IVUS images of the plurality of previous patients;
• and
• labeling the first reference IVUS images with stent area values of the second reference IVUS images, and
• wherein the training comprises:
• extracting, based on the first processed IVUS images and stent procedure information of the plurality of previous patients, reference feature information about the stent procedure information;
• obtaining, based on at least one of the first reference IVUS images and the first processed IVUS images, reference mask image information in which a blood vessel boundary and an inner wall boundary are distinguished from each other;
• and
• training the deep learning model by using the stent area values of the labeled first reference IVUS images and by using, as training data, the reference feature information and the reference mask image information.