top of page

▶ Representation learning

In the field of medical image processing, representation learning offers valuable insights into the hidden characteristics of data. It automatically discovers meaningful features or representations from raw inputs. By learning abstract representations, it enhances performance across a variety of tasks, including classification, regression, clustering, and generative modeling.

3.png
1.png
2.png

▶ Generative AI

Generative models learn the distribution of training images and can generate new images with similar characteristics. Common types of generative models include Generative Adversarial Networks (GANs), Variational Autoencoders (VAEs), and diffusion models. These approaches are useful for tasks such as image synthesis, style transfer, and data augmentation. Additionally, generative models can be used to simulate functional time series data.

figure.jpg

Park et al., NeuroImage, 2024.

https://github.com/CAMIN-neuro/GAN-MAT

image.png

Jang et al., in preparation.

419805532-41ebacab-60f2-4b02-92f7-5b4c818ebabd.gif
419805531-e6772f6e-a776-4a3f-afff-d4f079fef383.gif

Kim et al., in preparation.

▶ Segmentation

Segmentation and detection of brain lesions are critical challenges in clinical practice. To reduce false positives, it is essential to incorporate advanced feature extraction, feature selection, and classification techniques. Additionally, parcellating the whole brain into distinct regions or networks is another important area of research. We develop advanced methods to improve both brain lesion segmentation and brain parcellation.

netmodel.jpg
캡처.JPG
swin.png

Kim et al., MICCAI BraTS Challenge, 2024.

그림1.png
Office: (02841) 612A, Science & Engineering Library, Korea University, 145, Anam-ro, Seongbuk-gu, Seoul, Republic of Korea (Map)
Tel: (+82) 02-3290-5924
E-mail: boyongpark@korea.ac.kr
globalsymbol_eng2_large.gif
bottom of page