Research And Implementation Of Diagnostic Classification Method Of Thyroid Nodules

Thyroid disease is a common endocrine disease,which mainly as hyperthyroidism,hypothyroidism,thyroidit is and thyroid nodules.Thyroid nodule is one of the most serious threats to human health and the incidence rate has increased year by year.Thyroid nodule patients has left a large number of electronic medical records data in their treatment,in order to improve the status of clinical diagnos is of thyroid nodules,we need to mine the information implied in these data efficiently and accurately.In the traditional clinical diagnosis of thyroid nodules,doctors need to take ultrasound,blood tests,fine needle aspirat ion and other tests to determine the patient's benign and malignant properties,but even so,the accuracy of diagnostic results is still unsatisfied.On the other hand,the traditional machine learning algorithms show high error when training and forecasting the real medical dataset.The reason is that it does not take the specificity of medical data sets-sparseness and imbalance into account,so the results have a greater bias.In this paper,in order to reduce the unnecessary inspection process and improve the accuracy and efficiency of thyroid nodule ident ification,we propose a method for the ident ificat ion of thyroid nodules based on the characteristics of ultrasonography by improving the existing integrated learning model.We established a self-defined identif ication model of thyroid nodules,designed and implemented an assistant ident ificat ion system of thyroid nodules based on ultrasound data.Firstly,we analyze the clinical data of patients with thyroid nodules,such as the basic information,biochemical indexes and clinical diagnosis of patients,and we also research the correlation between the indexes and the clinical diagnosis,which can provide an important basis for the clinical treatment of thyroid nodules.Next,we structure the thyroid echocardiography data and extracted the effective and structured feature attributes,then do some works like balanc ing,digit izing and other necessary preprocessing to get the data which can be identified by the machine learning classif ication algorithm.It is convenient to the experimental process of data analysis and modeling.Finally,on the basis of the existing integrated learning model,a new authentication model is constructed by adding a custom item to the objective function.It can effectively solve the sparseness and unbalance of the dataset and improve the accuracy of forecast results.At the same time,a new assistant system of thyroid nodules based on ultrasound examinat ion data can be established to realize the automatic identificat ion function.By inputting the corresponding ultrasonic features,the patients and doctors can predict the identif ication result of thyroid nodules in real time which can improve the efficiency of examination.In order to verify the superiority of the method proposed,the experiment compares the algor ithm with random forest,support vector machine and neural network algorithm both on the real medical dataset and the UCI standard dataset respectively.The results show that the method has the highest accuracy rate and achieve 92.43% and 94% respectively.