Study On Image Enhancement Of Coronary Angiography Based On Improved Retinex-net Model

Coronary angiography is an important method for diagnosing cardiovascular diseases and has been widely used in clinical diagnosis.However,due to the limitation of imaging mechanism,there exists uneven gray scale area in the blood vessel that is caused by nonuniform distribution of contrast agents in the coronary angiography images.Meanwhile,uneven ray exposure and some tissue structure also interfere with the enhancement of blood vessel structure,which affects the doctor’s assessment of patient’s cardiovascular condition.Tradition alvascular image enhancement method enhances the blood vessel image,but it also increases the noise in the image,making the experimental results susceptible to noise interference and affecting medical diagnosis.To obtain clearer vascular images,the research on coronary angiography image enhancement was carried out in this paper.Based on the analysis of the traditional enhancement algorithm,it focused on the bilateral filtering,dark channel prior and Retinex principle,and further improved the Retinex-Net network model,thus forming a kind of more effective coronary angiography vascular enhancement algorithm.The main work of this paper is as follows.The bilateral filtering method was adopted to process and the decompose the vascular image.The detailed vascular information was retained through the pyramid structure to synthesize the low-illumination and normal light image,which would be used as a standard for vascular image training to carry out subsequent experiment.The improved Retinex-Net network model was used for vascular enhancement.Aiming at the problem of brighter areas in blood vessel images,the dark channel prior principle was applied to reduce the influence of brighter areas.By modifying the loss function of the enhanced network in the Retinex-Net network model and combining with the dark channel priori principle,the influence of brighter area on the model could be effectively compensated,thereby realizing the enhancement of blood vessels.In order to obtain the better detailed information of the blood vessel images,we add the reflection network to the Retinex-Net model to enhance the reflection image of the blood vessel.Selecting the multi-scale structural similarity loss function as the loss function of the network model can improve the model’s recognition accuracy of the blood vessels in the reflected image effectively,so as to achieve the experimental effect of blood vessel enhancement.Through comparison experiments,the results are analyzed and compared by using Retinex principle,dark channel principle and Retinex-Net.Experimental results show that the improved network model can effectively enhance the blood vessel part and achieve better visual effects.Compared with the experimental results using other loss functions,the modified loss function can effectively improve the vascular recognition degree,thus reducing the influence of noise on the vascular image.