基于频域状态空间模型和注意力增强的胃早癌检测OA

Gastric cancer remains a malignancy with high incidence and mortality rates worldwide.Early intervention in gastric cancer can improve patient survival rates,and gastroscopy is the most effective method for screening early gastric cancer.In re-cent years,the use of computer technology to assist medical image processing has rapidly advanced.To enhance the accuracy of early gastric cancer detection,this paper proposes an early gastric cancer detection network based on a Frequency domain State Space model and Attention Enhancement(FSSAE).This method enhances the frequency domain analysis of critical information in gastroscopic images through the Frequency domain State Space model(FSS),improving the recognition accuracy for early gas-tric cancer.Additionally,an Attention Enhancement Module(AEM)is incorporated to improve the network's ability to focus on cancerous regions.A new loss function is also introduced to further enhance network performance.Extensive experiments are con-ducted on the constructed early gastric cancer dataset using the proposed FSSAE network.Ablation experiments validate the ef-fectiveness of FSS and AEM.Comparative experimental results show that the FSSAE network outperforms baseline models and other networks,with the FSSAE's AP metric improving by 2.81%over the baseline network,increasing the detection accuracy for early-stage gastric cancer.