Research On Multispectral Non-invasive Detection Methods Of Blood Vessels

In the field of medical imaging and diagnosis,the existing vascular imaging technology can only realize single band shallow blood vessel imaging,and it is difficult to obtain threedimensional visual information.In this paper,a multispectral vascular non-invasive intelligent detection system is developed,which provides a solution for multispectral imaging of deep vessels and obtaining 3D parallax data.The main research work is listed as follows:(1)In the case of high cost of obtaining medical samples and their labeled data,aiming at the problems of the current traditional algorithm of blood vessel extraction,such as weak generalization ability in complex texture blood vessel data,too much dependence on data based on deep learning and insufficient timeliness of training reasoning,a super real-time blood vessel extraction model based on texture suppression and detail enhancement is established.In this model,perceptual weight module,detail coding module and hard sample screening cumulative constraint module are introduced to complete the module construction,and the end-to-end high generalization output is completed.Experimental results show that compared with other similar algorithms,this model has better extraction accuracy and timeliness.(2)In the application scene of weak texture vascular image,aiming at the problems of high mismatch rate of traditional vascular disparity matching technology,weak quantization ability of matching model based on deep learning and poor timeliness,a new algorithm of vascular disparity matching based on threshold truncation and integral mapping is proposed.The algorithm consists of three modules: threshold truncation,integral mapping,disparity calculation and complement calculation.The disparity result output of the improved gray step is obtained by morphological close operation.In the simulated blood vessel data set collected in this paper,the algorithm proposed in this chapter has higher accuracy.(3)On the premise of noninvasive and noninvasive detection,a multispectral noninvasive vascular detection system is built to solve the problems of single detection band,inconvenient use and high price of the existing vascular detection system.Aiming at the problem of high cost of vascular data integration,a simulated vascular data acquisition system is designed and built,which provides a cheap and reliable data base for quantitative analysis of algorithm.