Research On Breast Cancer Pathological Image Classification Based On Deep Learning

According to statistics,the potential risks of breast cancer to the lives and health of women is increasing day by day.In terms of clinical treatment,visual inspection of histopathology slides is one of the main methods used by pathologists to clinical assess the stage.This diagnosis method requires pathologists with rich clinical experience.It takes a lot of time to complete,and the diagnosis results are biased due to the shortage of medical resources and inevitable manual omissions.In recent years,deep learning has made considerable progress and progress in the field of image classification and recognition,and has also begun to emerge in the research of medical imaging,promoting the transformation of clinical pathological diagnosis from qualitative to quantitative.However,most of the existing pathological image classification research methods are based on traditional machine learning algorithms.These methods use manual extraction methods to extract image features.There are still many shortcomings and challenges in processing high-resolution pathological image slice data.Under this training mode,the efficiency of the model is generally low and it does not have good adaptability.In order to further promote clinical application and help doctors improve the consistency and efficiency of breast cancer pathological diagnosis,In this paper,combines the latest research results in the field of imaging to study the classification of breast cancer pathology images.The main research contents are as follows:(1)The pathological images with high resolution and rich details are difficult to be directly used as the input data of deep learning model.In this paper,the image is magnified proportionally,the image is sliced according to the input size of the model,and the unqualified images are removed according to the actual image coverage after the cutting,finally a series of sample images can be used for model training.(2)In order to solve the problem of breast cancer pathological image classification,This paper proposes an image classification method based on an improved Inception-V3 model on the basis of extensive research on deep learning technology in related fields and the current research status at home and abroad.Through the improvement of model structure,the introduction of parameter regularization method,and the adjustment of key parameters of the model,the overall framework of the network is further determined.The validity and feasibility of the model in pathological image classification are verified by a large number of contrast experiments.(3)Aiming at the problems of pathological image data,such as small amount and over fitting,based on the research content of the previous chapter,continue to use the transfer learning method to transfer the mature and weight information obtained from the massive data to the target architecture to achieve network model optimization process,and realize breast cancer classification task through network fine-tuning method.The experimental results show that in comparison with the convergence speed of the classic deep learning models and the accuracy of the results,the work done in this paper far exceeds the traditional deep learning methods,and effectively improves the analysis and treatment process of breast cancer diagnosis,and laying a theoretical and practical foundation for further advancement of clinical applications.(4)Aiming at the scarcity of deep learning assistant systems that can be used in the clinical field,and verifying the effectiveness of the proposed method,based on the improved deep learning models proposed in the previous paper,a classification and detection system for breast cancer pathological images is designed and implemented.The system intuitively displays the classification accuracy and rapidity of the improved model through user’s actual operation,and makes an exploratory attempt for future clinical application.