Advances in Machine Learning: New Findings in Lesion Segmentation and Artificial Intelligence

Strategic Insights -

Researchers at the School of Information Engineering have made significant progress in the field of machine learning, specifically in the area of lesion segmentation and artificial intelligence. The study, published in the International Journal of Pattern Recognition and Artificial Intelligence, presents a novel approach to segmenting lesions in medical images using the PST-UNet model. This model combines transformer and U-shaped structures to preserve spatial features of medical lesions, thereby improving the accuracy of medical lesion segmentation.

Key Takeaways:

  • The PST-UNet model combines transformer and U-shaped structures to integrate the encoder's multi-scale features, effectively preserving spatial features of medical lesions.
  • The encoder includes the Swin transformer block and the entire Gaussian Error Linear Unit (GELU) activation function, while the decoder uses the Swin transformer block, upsampling, and skip connections from the cascaded convolution fusion modules.
  • The research concluded that the PST-UNet model can improve the segmentation accuracy of normally distributed medical lesion data, which has a positive effect on the treatment of kidney disease.
  • The study achieved a 0.02289 improvement in accuracy using the PST-UNet model, showcasing its effectiveness in lesion segmentation.
  • The research was funded by the Science and Technology Plan Project of Huainan, Anhui Province Research Planning Project for the Year 2023: Sentiment Analysis of Academic Warning Comments among College Students Based on the CNN-BiLSTM-Attention Model, Application Research of Connective Origami Probe Machine.
  • The study has been peer-reviewed and published in the International Journal of Pattern Recognition and Artificial Intelligence.

Statistics:

  • The PST-UNet model achieved a 0.02289 improvement in accuracy.
  • The research was funded by the Science and Technology Plan Project of Huainan, Anhui Province.
  • The study was published in the International Journal of Pattern Recognition and Artificial Intelligence.
  • The research involved a team of authors from the School of Information Engineering, including Xuemei Shi, Xiedong Song, Ming Deng, Dalei Zhang, Xiaoyan Li, and Baoguo Chen.

Sources:

  • "Unet and Swin Transformer Fusion Network for Lesion Segmentation In Biological Kidney Imaging." International Journal of Pattern Recognition and Artificial Intelligence, 2025.
  • Xuemei Shi, Huainan Union Univ, School of Information Engineering, Huainan 232001, Anhui, People's Republic of China.
  • "Recent Findings from School of Information Engineering Provides New Insights into Pattern Recognition and Artificial Intelligence (Unet and Swin Transformer Fusion Network for Lesion Segmentation In Biological Kidney Imaging)." Robotics & Machine Learning, August 25, 2025, p 284.