Detection Of Lumbar Vertebrae Hyperplasia Based On Convolutional Neural Network

In recent years,artificial intelligence technology has developed rapidly,and machine learning technology has been continuously improved and developed.Deep learning has not only been widely applied in the fields of finance,agriculture,transportation,etc.,but also research in the medical field has gradually increased.In the medical field,deep learning He has made outstanding achievements in lesion segmentation,disease diagnosis,and image classification.In the research of cancer,brain tumor,heart disease,skin disease,Parkinson’s disease,epilepsy,depression and other diseases,fruitful research results have been achieved,which not only eased the work pressure of medical staff,but also improved the technical level of medical equipment.Improve the diagnostic efficiency of the disease and facilitate the timely medical treatment of patients.However,there are currently few studies on artificial intelligence in lumbar disease.This topic is aimed at the research trend of artificial intelligence in the field of medical imaging.Taking lumbar vertebrae hyperplasia as the research object and based on the lumbar X-ray image,this paper proposes a classification of lumbar vertebrae hyperplasia based on convolutional neural network.The neural network model realizes the intelligent detection classification of lumbar vertebrae joint hyperosteogeny,and identifies the lumbar vertebra joint image input into the neural network model to determine whether there is bone hyperplasia.In the course of the study,a lumbar vertebrae hyperplasia image dataset was established,including 4245 images of the lumbar vertebrae frontal image,including two categories:positive disease,positive normal,and four types of positive and lateral lumbar vertebrae images.Zhang,including four categories of positive disease,normal frontal,side disease,and normal side.In the lumbar image data,image preprocessing is performed on the collected lumbar image,including screening,intercepting,amplifying,naming,sizing,and image edge extraction.Based on the layer design of LeNet network model and the parameter design method of AlexNet network model,this paper combines two networks and proposes a small and lightweight convolutional neural network model,which can be used for small-sized lumbar vertebrae.Accelerated images are identified and classified.The neural network model is fine-tuned by adjusting the number of convolution kernels,the size of the convolution kernel,the learning rate,and the parameters of Dropout.The network model is gradually optimized,and an efficient method is obtained in the process of continuously training the model.Identify accurate network models.Under the Linux system,based on the development framework of TensorFIow,the Python programming language was used to classify the bone proliferative diseases of the lumbar vertebrae joints,and the two-class experiment of the lumbar vertebrae frontal joint images was performed.In the two-category experiment,the test set The recognition accuracy of the above classification reached 93.29%,and the area AUC value under the ROC curve reached 0.97.In the four-category experiment of lumbar vertebrae and lateral vertebra joint images,the classification accuracy on the test set reached 87.95%;the lumbar vertebrae joint The images were input into several neural network models,and the classification results were compared.The accuracy of convolutional neural network model classification was higher than 86.8%of LeNet network model.83.76%of AlexNet model,VGG-16 network model.The classification accuracy rate of 90.34%.Relying on artificial intelligence technology,the work pressure of medical staff is greatly reduced,the misdiagnosis rate of clinical diagnosis is reduced,the patient is effectively treated in time,the delay of the disease is avoided,and the diagnosis of lumbar hyperosteogeny disease is provided more scientifically.The basis for diagnosis.