Super Resolution Algorithm Of Diffusion Weighted Magnetic Resonance Image

Diffusion weighted magnetic resonance imaging(DW-MRI)is the only noninvasive in vivo imaging technique to detect the movement of water molecules in living tissues.It can provide tissue structure information that is different from ordinary magnetic resonance images,and can reveal the microstructure of neurons and the connectivity of brain white matter fibers.In the diagnosis and treatment of brain diseases,it can provide unique information for the survival and development of the brain.As a potential and valuable imaging technology,it has gradually become the first choice of acute brain injury examination and has been broadly applied to more clinical diagnostic findings.However,due to the limitations of imaging equipment and environment,the resolution of DW-MRI is usually limited.Low quality image will affect the doctor’s analysis and diagnosis of the patient’s condition,and the information extracted from the image is very important for the doctor’s diagnosis.From the clinical and researcher’s point of view,high quality image is very necessary.Therefore,it has become an important research topic in the field to improve the quality of scanned image through effective and reliable processing methods.Generally speaking,the higher the magnetic field intensity and the longer the scanning time,the higher the image quality.However,due to the influence of the actual environmental conditions and costs,it is often unable to obtain the images that meet the actual needs.Therefore,using post-processing method to improve the resolution of the image is very desirable,instead of depending on imaging technology.To solve this problem,this paper studies the super-resolution algorithm to improve the image quality of DW-MRI and restore more details.The main contents of this article are structured as follows:(1)A super-resolution algorithm based on auto-encoder is proposed.The framework can effectively extract key features from massive data and feed them to the network for training,which deeply enhanced the computational efficiency.Fusion of ultra-high magnetic field prior feature information,through the convolution kernel size set by the network automatically extract the image’s own feature information,so that the results have more reconstruction details.Experimental results show that the proposed algorithm can recover more information in the process of super-resolution reconstruction and has better perception quality.(2)Based on the generative adversarial network,an improved generative adversarial network model is proposed.Scale specific processing module is used to deal with scaling and image quality reconstruction in multi-scale.The sequential attention mechanism is designed,which combines time domain and spatial information to intuitively imitate the way of doctor’s examination data from the perspective of clinical application.The information of the current slice is fully perceived by rolling up and down.Through the optimization of the objective function and experimental verification,the proposed model has good effect and robustness.Based on deep learning technology,this paper transfers the rich information collected by expensive medical imaging equipment which has not been widely used in clinical practice to the low-quality data of clinical practice.The anatomical prior information of ultra-high magnetic field is learned in order to achieve better results when performing super-resolution tasks in ordinary low magnetic field.Aiming at the existing image quality problems of DW-MRI,this paper investigates the research status at home and abroad,and analyzes the data characteristics.On this basis,the research work is carried out and two algorithms are proposed.The reconstructed DW-MRI image quality has been improved,which is conducive to further accurate diagnosis of doctors,and has certain research significance and clinical application value.