Recently, Yan Shun, a 2023-level master’s student in the Mechanical Engineering program, has published a research paper titled “Multi-scale Convolutional Attention Frequency-enhanced Transformer Network for Medical Image Segmentation” as the first author in Information Fusion (a top journal in the first zone of the Chinese Academy of Sciences, IF: 14.7). Taizhou University is the sole affiliation.

The paper addresses the shortcomings of traditional Transformers in extracting local features and retaining detailed information by proposing a multi-scale convolutional attention frequency-enhanced Transformer network combined with wavelet transform. It aims to utilize wavelet transform to retain time-frequency features of images at different scales, effectively capturing high-frequency information such as texture and edge details to improve the accuracy of image segmentation.

Figure 1: Multi-Scale Convolutional Attention Frequency Enhanced Transformer Network

Figure 2 Grouped Separable Aggregation Convolution Module (GSACM) and Efficient Frequency-Enhanced Transformer Module (EFTM)

This research achievement proposed a novel Transformer network model applicable to medical and health scenarios such as disease diagnosis, intelligent image analysis, and smart healthcare. It was funded by the National Natural Science Foundation of China under the project “Multimodal Personality Recognition Based on the Fusion of Auditory and Visual Information Using Generative Adversarial Networks.”

Paper Link: https://doi.org/10.1016/j.inffus.2025.103019