Research on Multi-modal Medical Image Fusion Based on Multi-Scale Attention Mechanism

Abstract

Image fusion technology has been widely used and researched in the fields of medicine, aviation, military and civil. Compared with the traditional image fusion technology, the fusion image technology based on intelligent algorithm generates more realistic, clear and reliable images, and has a lot of detailed information. Deep neural networks have rapidly developed into a research hotspot in medical image analysis, which can automatically extract hidden pathological information from medical image data. Multi-scale and attention mechanism are two important modules in neural networks, which can significantly enhance the features extracted by the network. The problems that need to be solved for medical image fusion based on multi-scale attention mechanism include: how to build a multi-scale module, how to build an attention module, and how to combine multi-scale and attention mechanisms. Construct a deep learning network through a multi-scale attention mechanism, and then adjust the parameters and train the network to achieve multi-modal medical image fusion. In the process of building neural networks, multi-scale and attention mechanisms are incorporated, and multi-scale features of multi-modal medical images are extracted and enhanced for fusion. Through a large number of experiments: (1) The edge strength of the fused image is expected to increase by 10%-20% based on the average value of the existing algorithm; (2) The fused image can achieve high color fidelity and rich detailed information; (3) ) The time required for the fusion algorithm is expected to be reduced by 1%-10% based on the average value of the existing algorithm.