| Maternal and child health is an important part of the medical and health field.The health of women and children is the cornerstone of national health.China also attaches great importance to the maternal and child health work.With the rapid development and wide application of computer and Internet technology,the informationization of maternal and child health of China has achieved remarkable results,and a large and rich variety of data has been collected.Analyzing and mining valuable information stored in data has become a major focus of follow-up work.This thesis selects X-ray of hand bone and structured disease data as the research focus,analyzes and mines the information stored in them,to provide support for the development of maternal and child health work.The human bones grow with a degree of regularity.Bone age is a evaluation standard of the level of children’s physiological development.Bone age assessment helps to prevent and detect some pediatric diseases in time.This thesis investigates the application of convolution neural network and attention mechanism to the task of bone age assessment on X-ray photographs of children’s left hand bones.Using ResNet-50,a classical convolution neural network model,as the image feature extractor,the effects of gender input,background noise of pictures,and image contrast on the accuracy of bone age assessment tasks in network model processing are studied step by step.A attention module composed of convolution layers and local connection layers is proposed to improve the performance of network model,and a comparison experiment is designed to verify the effectiveness of this module.Analysis and mining of related factors of diseases can help understand the causes of disease and prevent its occurrence.However,the imbalance of disease data has limited the effectiveness of traditional association rule mining algorithms which is based on support and confidence,resulting in the omission of important information.To solve this problem,this thesis proposed a correlation factor analysis algorithm based on relative risk and tests on the dataset of congenital heart disease in children.This thesis preprocesses the congenital heart disease datasets with missing value filling and discretization of numeric attributes,introduces correlation factor analysis algorithm based on relative risk and corresponding two pruning methods,designs comparison experiments to verify the effectiveness of the algorithm on unbalanced dataset mining and the redundancy reduction of the two pruning methods,and use the odds ratio to verify the mining results again.Combining the above two algorithms,a B/S architecture of maternal and child health system is designed and implemented.The system has functions of data collection and preprocessing,bone age assessment,height prediction,disease analysis and so on.This thesis introduces the detailed function points,interactive design,database design and the implementation of the background interface of the system,and shows the effect diagram and test report of the implemented system. |
