Zhiwen Yang

Zhiwen Yang

I have a wide range of research interests, including machine learning, deep learning and their application to computational biology, with a focus on developing tools to solve real world problems.

MedSegKAN: A superior medical image segmentation method based on the improved KAN structure

Published in Proceedings of the 16th International Conference on Graphics and Image Processing (ICGIP 2024), 2024

This project addresses the accuracy limitations of traditional KAN architectures in medical imaging.

Key contributions include:

  • Employed Gaussian smoothing preprocessing, boundary loss functions and regularization strategies to reduce segmentation errors.
  • Incorporated the ECA Attention Module to enhance feature focusing capabilities.
  • Achieved a Dice score of 92.89% in medical image segmentation tasks, highlighting the high efficiency of the method.

Recommended citation: Fang, Z., Yang, Z., Zhang, X., & Han, Q. (2024). MedSegKAN: A superior medical image segmentation method based on the improved KAN structure， In Proceedings of the 16th International Conference on Graphics and Image Processing (ICGIP).
Download Paper