Breakthrough in Gastric Cancer Detection: A Novel Image Segmentation Approach

A team of researchers at Nanjing University has made significant strides in the early detection of gastric cancer, a leading cause of cancer-related deaths globally. By leveraging advanced microscopic hyperspectral imaging technology and a novel image segmentation approach, the researchers have demonstrated improved accuracy in identifying precancerous lesions. The study, published in Optics and Laser Technology, presents a multi-scale feature fusion and attention-enhanced U-Net (MSFA-Net) algorithm that effectively addresses the challenges in segmenting complex lesion areas.

Key Takeaways:

• The research team, led by Qi Zhao, developed a novel image segmentation approach called MSFA-Net, which uses multi-scale feature fusion and attention-enhanced U-Net to improve early gastric cancer detection.
• MSFA-Net effectively addresses the semantic gap between the encoder and decoder, enabling the accurate identification of cancerous regions.
• The algorithm demonstrated promising results in segmenting precancerous lesions in gastric cancer using microscopic hyperspectral pathology images.
• The Cross-Stage Feature Fusion (CSFF) module and Multi-Scale Integrated Convolution (MSIConv) were used to accurately merge encoder-level features and extract multi-scale features, respectively.
• The Adaptive-weighted Attention (AWA) module was designed to optimize the fusion of encoder and decoder features, enhancing the recovery of image details.
• Experimental results showed that MSFA-Net performed well on both Intestinal Metaplasia (IM) and Gastric Intraepithelial Neoplasia (GIN) stages of precancerous gastric cancer.

Statistics:

• The research was funded by the Nanjing Health Science and Technology Development Special Fund Project.
• The study used microscopic hyperspectral pathology images to segment precancerous lesions in gastric cancer.
• MSFA-Net demonstrated improved accuracy in identifying cancerous regions compared to traditional segmentation methods.
• 187 is the issue number of Optics and Laser Technology where the study was published.

Sources:

• NewsRx. Research Conducted at Nanjing University Has Updated Our Knowledge about Gastric Cancer (Msfa-net: Multi-scale Feature Aggregation and Attention-enhanced U-net for Microscopic Hyperspectral Pathology Images Segmentation). Health & Medicine Week. September 5, 2025; p 4984.
• Qi Zhao, et al. Msfa-net: Multi-scale Feature Aggregation and Attention-enhanced U-net for Microscopic Hyperspectral Pathology Images Segmentation. Optics and Laser Technology, 2025;187.