Research On Medical Data Classification Algorithm Based On Machine Learning

Genetic data and medical imaging data are two kinds of representative medical data.This thesis embarks from the two kinds of medical data.Machine learning model was built to classify clear cell renal cell carcinoma.Deep learning model was constructed for diabetic retinopathy data classification.Focusing on feature extraction and model reconstruction,this paper discusses the problems and solutions of machine learning and deep learning in the diagnosis of major diseases.The algorithm and model are redesigned to improve the accuracy of diagnosis.The classification performance of the target data sets is better than the existing optimal methodClear cell renal cell carcinoma is the most common subtype of renal cell carcinoma Gene detection technology can be used to understand cancer at the molecular level and obtain genetic data.On the basis of second-generation gene sequencing technology,machine learning can be used for computer-aided diagnosis of gene data.However,when machine learning classification algorithm is applied to genetic data,there are problems such as large data noise,small number of data samples,high feature dimension and strong correlationDiabetic retinopathy is the most important form of diabetic microvascular disease It is a special fundus lesion and one of the complications of diabetesThere are many problems in the diagnosis of diabetic retinopathy,such as high resolution of fundus image,small pathological features,data imbalance,and small amount of data,etc.The traditional deep learning model is not suitable for such high-resolution data as fundus image of diabetic retinopathy.Based on the above background and problems,the following work is accomplished in this paper:1.A new classification model of clear renal cell carcinoma based on gene data and machine learning was proposed.This paper improves the pretreatment method of gene data,improves the stability of gene expression profile data and the robustness of classification model through box coding,and proposes a joint gene selection algorithm to extract key gene characteristics.This classification model based on genetic data and machine learning can extract fewer features in the classification of clear renal cell carcinoma tumor samples and achieve higher prediction accuracy and stability than the existing optimal model2.A diagnosis model of diabetic retinopathy based on deep learning was proposed The method proposed in this paper combines data enhancement and resampling to quantitatively analyze the complexity of network models with different input scales Finally,a deep neural network model suitable for high-resolution images of sugar network is designed by using the strategy of gradual fine-tuning training.The model performed better in EyePACS data than the Kaggle sugar net lesion detection champion method,and significantly improved detection of mild lesions.In a word,this paper systematically demonstrates the application of machine learning in pathological image analysis and genome data analysis to realize disease diagnosis and prediction,which is of great significance for assisting clinical decision making and providing intelligent and personalized medical services.