Research On The Diagnosis Of Hypertension Based On Deep Learning

The application of artificial intelligence in the field of medical imaging is a research hotspot in the computer vision.In recent years,artificial intelligence has developed rapidly in various fields,and smart medical care is the foundation for the development of smart hospitals in the future.In the field of medical image,such as X-ray,ultrasound imaging,etc.,the development of artificial intelligence in the field of medical imaging has become very important because of the difficulty of doctor reading and the high workload of young doctors.Hypertension is a disease with a high incidence in the world,and its serious complications may even threaten the lives of patients.The pathological basis of hypertension is the changes in the structure and function of blood vessels.The earliest changes are some tiny blood vessels,mainly in the eyes(fundus retinal blood vessels)and glomerular capillaries.In recent years,an optical coherence tomography angiography(OCTA)has emerged,which can observe relatively small vascular structures and read relevant vascular information from pictures.In order to solve the complexity of hypertension on OCTA images and the problem of small samples,this article will study the problem through two ideas.The first is to use the text feature data generated by the OCTA instrument to solve the problem and classify the hypertension based on useful features;the second idea is to train the image and use the deep learning method to classify the hypertension.By comparing the two ideas,the image-based model training effect is better than the textbased model training effect,and the Swin Transformer network is selected as the baseline model for improvement,and the mid-term fusion and late fusion methods in the multimodal fusion method are used for the macular area The OCTA image and the OCTA image of the visual panel are fused,and finally the late fusion model Multi-Swin and the mid-term fusion model SEFMST are established.The comparison shows that the SEF-MST effect is more superior,and the average accuracy rate of the five-fold cross-validation reached 85.30%.,Which is higher than the 77.40%accuracy rate obtained by the doctor through the Keith-Wagener-Barker standard diagnosis..After getting a better model,we built a multi-functional diagnostic platform based on ophthalmic OCTA images through Python Flask module and Javascript language.The platform mainly has ophthalmic hypertension diagnosis functions,ophthalmic blood vessel segmentation functions,image annotation functions and medical records.Upload function.Through the above-mentioned functions,this platform will finally be built into a disease diagnosis platform suitable for ophthalmic OCTA images,a high-quality OCTA data set construction platform and a social-oriented data sharing platform,which will lay the foundation for the future development of the OCTA platform.Finally,the research content of this article is summarized,and the future research work is prospected.