Research And Realization Of Benign And Malignant Thyroid Nodule Prediction

With the vigorous development of data mining technology,data mining shine in all areas.The application of data mining technology in the medical field for disease prediction has become a hotspot of the current research.At the same time the hospital in the provision of medical services for patients in the process has accumulated a large amount of valuable clinical data resources,more and more clinical data is stored in the medical database,how to discover valuable information from a large number of irregular clinical data and contribute to the future of disease diagnosis and medical research become a hot topic in the era of large data.Deep learning technology is the latest research in the field of machine learning,the paper attempts to apply the deep learning technology to the thyroid nodules of benign and malignant prediction.In this paper,a hypothalamic thyroid nodule ultrasonographic data mining method was used to establish a predictive model based on deep learning and to design and implement a benign and malignant prognostic system for thyroid nodules in a large comprehensive tertiary hospital in Shanghai.The main work of the thesis is as follows:(1)Multi-dimensional analysis of thyroid disease diagnosis.Dimensional analysis and visualization.In order to explore the inherent laws and characteristics of research data,we hope to analyze the patients with thyroid diseases by multidimensional analysis,and visualize the results in a graphical way.(2)To establish a benign and malignant prediction model of thyroid nodules based on Deep Belief Network(DBN).First,ultrasound data of the thyroid nodule is obtained in the form of a doctor's diagnosis,and is mainly composed of text data.So it is necessary before the model to the data preprocessing.Pre-processing is mainly text structure,data cleaning,numerical and normalization.Then we construct the model of benign and malignant thyroid nodule prediction.Combined with the characteristics of thyroid nodules ultrasound data,based on the construction of benign and malignant thyroid nodules DBN prediction model.The training core of this model includes RBM and BP(Back Propagation)neural network.The algorithm is trained by the layer-by-layer greedy learning algorithm.The optimal network structure is determined by comparing several experiments,and a neural network model based on DBN of benign and malignant thyroid nodules prediction model.(3)To design and implement the prototype system of benign and malignant thyroid nodule prediction.The model of benign and malignant thyroid nodule prediction system was designed and realized by using DBN model.The system consists of four modules: data preprocessing module,multidimensional analysis module,prediction model module and prediction result display module,which provides a method for predicting the benign and malignant thyroid nodules.