Application Research Of Deep Learning In Brain Tumors Image Recognition

Brain tumors are one of the most common tumors that threaten the safety of human life.The traditional diagnosis of brain tumors is based on the adverse symptoms of the patient’s body and the complications of the fundus.With the development of medical technology,the diagnosis of brain tumors now has changed to be judged by doctors based on MRI images of the patient’s brain,but with the daily workload of doctors and the different clinical experience of each doctor,misdiagnosis and missed diagnosis occur in long-term clinical diagnosis.Computer aided diagnosis system(CAD)can identify images accurately,assist doctors in daily work,relieve working pressure and improve accuracy of diagnosis.At present,there are few computer-aided diagnosis system and related research about brain tumors.This paper proposes an Inception-CNN model based on improved Inception modules,aiming at identification and classification of three major brain tumor disease.Through training and testing,the results show that the accuracy of the model is up to the standard that can assist doctors in making diagnostic decisions.The work done in this paper is as follows:1.Brain tumor MRI image dataset is screened and processed,in view of the problem of uneven data set category,different image enhancement methods are used to expand the data set and compare the effect of several data enhancement methods.Choose the best way to expand and balance the data set used.2.The Inception-CNN model based on improved Inception modules is constructed to identify and classify Kaggle Brain Tumor MRI Image Dataset.Experiments show that the model can complete the task of recognition and classification of brain tumor MRI images more accurately and effectively,it can meet the standard of daily diagnosis for auxiliary doctors.3.Attention mechanism is introduced into the Inception-CNN model to overcome the problem of unmarked focus area of the dataset.By testing on the Kaggle Brain Tumor MRI Image Dataset,it is proved that the addition of attention module has improved the performance of the Inception-CNN model.By testing on the CE-MRI dataset,it is proved that the Inception-CNN model combined with attention module can complete the recognition and classification task of the data set with higher accuracy.