Prediction Model Of Coronary Heart Disease Based On Multi-source Feature Analysis

In recent years,the incidence of cardiovascular disease continues to increase,and mortality is the highest among all diseases.Coronary atherosclerotic heart disease,is one of the most common types of cardiovascular disease.Although the treatment technology for coronary heart disease has made considerable progress,with the gradual increase in the incidence rate,the means of clinical treatment can no longer satisfy the needs.Therefore,establishing an effective predictive mechanism and advance intervention for diseases is necessary.However,the general cost of the existing risk prediction model has a relatively large cost.It requires not only the collection of long-term index data for the patient,but also long-term follow-up and return visits for patients.The period is more than 2 years,or even 5-10 years.Other factors,e.g.,the geographical,growth environment,are also limited.The time and space span is too long,which will inevitably affect the quality of data collection,and thus affect the prediction effect of the model.For the above reasons,it is very necessary for medical institutions to create a new type of risk prediction program to make up for the deficiencies of the original forecast model.With the support of medical related theory,this article uses ear characteristics to establish an image-based coronary heart disease risk prediction model.The overall structure of this paper is mainly organized as three parts.First,the Haar feature of face recognition is used to achieve a complete segmentation of the ear image.Second,the ear image features are constructed to make the prediction of coronary heart disease.The specific features include: the study of ear distance features with genetic information;depth mining of the neural network features associated with the coronary heart disease;and the analysis of the pathological features of abnormal texture in the ear region.Third,two different architectures of the algorithm model framework are established.The best prediction model structure is explored from the two directions of feature fusion and integrated learning.Finally,the overall evaluation of the performance of the prediction model is completed through the comprehensive comparison of various algorithms.The research work of this project is still continuously deepening and improving.It is expected that with the gradual accumulation of sample data and further exploration of ear feature points,the overall performance of the prediction model will be greatly improved.