Research On Intelligent Classification Of Idiopathic Scoliosis Based On Heterogeneous Dat

The application of machine learning in the medical field is developing rapidly.In some fields,its effect can even be comparable to that of professional doctors.In spinal surgery,machine learning models can help doctors to analyze the biological data and images of the spine.Many of them are used to assist diagnosis and treatment,but there are no landmark results.The automatic classification of diseases based on heterogeneous medical records is an important research topic in intelligent diagnosis.Based on the biological data and X-ray images,the diagnosis of AIS is studied in this thesis.Aiming at the two main problems of small sample and class imbalance,our methods achieve good effect by improving the existing machine learning classifiers.The main innovations are as follows:(1)An integrated model IMLBoost is proposed to deal with the imbalance of small sample datasets,which is applied to the spine biological data,and its innovation lies in: a cost-balanced loss function is designed and embedded into the ensemble model.It can assign befitting costs to different samples,taking the influence of minor samples,hard samples,and boundary samples into account,so it overcomes the shortcomings of traditional machine learning models in dealing with imbalance.(2)A novel capsule network IMLCapsNet is proposed for the classification of spine X-ray images,in which the SE module is introduced to learn important local features.Then the proposed loss function is used for the optimization of the training.Experimental results show that the model can deal with the imbalance and improve the classification accuracy of small sample data effectively.(3)An improved convolutional network IMLNet is developed for the classification of images.Some MBConv blocks of EfficientNet are used for the extraction of primary features,and then the improved RFB module is used for fusing features with different scales.Finally,by the proposed loss function,the model achieves good results on the data collected.In conclusion,this thesis demonstrates the application of machine learning in the classification for X-ray images and biological data,and realizes the diagnosis and prediction of AIS.It is of great significance for assisting clinical decision-making.