댓글 0건 조회 59회 작성일 2023-09-16 14:06
- Professor Sang Hyun Park's team developed a domain-adaptive artificial intelligence model technology specialized in medical imaging.
- It is expected to contribute to the development of deep learning models for new forms of medical images, reducing costs and time, and greatly contributing to the development of diagnostic software that can be used universally in various medical sites.
- The proposed technique extracts texture information from the structure information of the image and the image of the new domain, and generates a realistic image using the discriminator error function. In addition, the texture information of the new domain is maintained using the texture co-occurrence error function, and the mutual information error function is used to generate an image with less structural deformation. And based on this, a deep learning model is learned to perform domain adaptation.
- Using the proposed technique, domain adaptation is performed by performing zoning on fundus images and MRI data collected by various institutions, and by generating images of different styles in reverse for heart zoning for CT and MRI. This study is published in Pattern Recognition (IF=8.0), the top international academic journal in the field of artificial intelligence.