Research on the Optimization and Visualization of Biostatistical Method Based on Improved Residual Network

This paper proposes an image recognition method of biological cancer cells based on improved Unet model to assist doctors in carrying out more accurate disease diagnosis. Among them, Unet model is taken as the basic image recognition method, and the recognition performance of the model is further improved by adding attention module and residual module. The experimental results show that compared with other image recognition methods, the designed Unet model based on mixed attention mechanism and residual module has higher recognition accuracy, and the recognized images are more close to the real image, with MPA, Miou, and Dice of 0.86, 0.73, and 0.85, respectively. Therefore, the designed image recognition method of biological cancer cells based on improved Unet model has high recognition accuracy, can be applied to actual diagnosis and treatment scenarios, and can assist doctors in conducting more accurate disease diagnosis, which has high feasibility.