Research On Transfer Learning Classification Method For Medical Data

In recent years,with the development and wide application of artificial intelligence and machine learning technology,the field of medicine has undergone major changes and entered the era of digital diagnosis.A large amount of biomedical data is presented in the form of charts and images,such as MRI,EEG,and ultrasound imaging.Facing the challenge of growing data volume,automatic processing and identification of medical data has become an urgent problem to be solved.Due to the complexity and redundancy of medical data,researchers use machine learning methods to assist doctors in disease analysis and improve diagnostic efficiency.EEG is one of the most important diagnostic and therapeutic tools in the treatment of epilepsy.However,due to the large variance of EEG data,its distribution pattern varies significantly among different subjects,and the simultaneous collection and annotation of EEG data also requires a lot of time and human resources,resulting in a very large amount of available data.limited.To solve these problems,we use machine learning methods to automatically identify and diagnose epileptic EEG signals and assist doctors in EEG classification tasks.The specific work of this research is as follows:1)For the identification of epilepsy EEG signals,a classification method SDA-SSL-TSK-FS based on transfer learning and semi-supervised learning is proposed.The new method takes the TSK fuzzy system as the framework,which makes the decision results of the model interpretable.Using transfer learning to adjust the data distribution between related domains,the problem of model performance degradation caused by large spatial feature difference is improved.At the same time,semi-supervised learning is used to make full use of existing tags to obtain more hidden space feature information through membership function similar to FCM clustering idea,which further improves the learning ability of the model.2)In view of the unstable distribution of epileptic EEG data,the modular domain adaptive method PDA-ELM is used for automatic diagnosis.The model is based on the extreme learning machine,and the optimal output weight is obtained by the least square method.The model can construct independent and stable data sets,and then the classifier can learn the optimized data features,and finally make decisions.The learning rate of this method is fast,and it can quickly respond to the changes of EEG signals.Through the domain adaptive transfer learning,the extreme learning machine can be applied to more complex epileptic EEG recognition scenes,and the generalization ability of the model is improved.