| Worldwide,early breast cancer is difficult to be discovered because of its uncertain pathogenesis and no obvious symptoms.With the development of society,the number of patients with breast cancer has increased dramatically and the mortality rate is extremely high.Early breast cancer the probability of successful cure of cancer is very high.Early diagnosis can not only improve the survival rate of patients but also reduce the risk of prognosis for patients which has great significance.In medicine,Dynamic Contrast-enhanced magnetic resonance imaging(DCE-MRI)is widely used.It is used for early detection of breast cancer.However,this method not only requires excessive professional knowledge,but also requires a lot of human resources,and there is also a risk of subjective judgment.With the development of imaging technology and deep learning technology,intelligent medical image recognition can process a large amount of data,To achieve medically assisted diagnosis under the condition of ensuring extremely high recognition accuracy,and to provide objective diagnosis results for the diagnostic personnel,it has high clinical value for the recognition of breast cancer pathological images.To this end,the research of this thesis is based on the GoogLeNet model’s deep learning method to automatically classify breast cancer pathology images,and explore an improved convolutional neural network model to improve the accuracy of classification.This paper establishes the depth based on GoogLeNet.Convolutional neural network architecture model,using Softmax classifier to achieve feature recognition,and using migration-fine tuning method to prevent data over-fitting problem and data enhancement method in training model to expand data set and increase training scale.Experimental results show that in this paper the diagnostic accuracy of the method used reaches 90%.It can provide reference for the fusion of deep learning image recognition technology and modern medical imaging diagnosis. |
PDF Full Text Request