Research On Classification And Segmentation Of Breast Cancer Pathological Image Algorithm Based On Deep Learning

Pathological image analysis of breast cancer,a common malignancy among women,is an essential part of the diagnosis process.However,pathological images carry rich information,which requires a lot of time and energy of pathologists to ensure accurate diagnosis,and The danger of misdiagnosis is considerable.With the ongoing advancement of science and technology,can drastically enhance diagnosis accuracy and efficiency,decrease misdiagnosis rates,and provide improved medical care for breast cancer sufferers.The specific work of the paper is as follows:(1)A bilinear structure-based deep learning network model SE-Bi Pool Net is suggested for the precise categorization of breast cancer pathology images.Adopt image transformation and image enhancement technology to collect more pathological images and alleviate the problem of missing pathological images.At the same time,in view of the exposure problem caused by the poor image quality of histopathological slice data,a chroma-preserving brightness enhancement(CRLE)method is proposed.SE-Bi Pool Net introduces SE-Bi Attention module and Dual Channel Pooling module.SE-Bi Attention module can adaptively adjust the weight between different feature maps to better capture the relationship between features;Dual Channel Pooling module can extract features from two channels,and obtain richer feature information through interactive operation,thus improving the network’s perception of details.The model’s efficacy in classifying pathological images has been demonstrated by experimental results.(2)A DFU-Net image segmentation model based on single-path attention-double path parallel convolution is proposed.Iterative threshold segmentation and adaptive binarization,combined,are employed to eradicate the background data of pathological images and effectively distinguish the areas of concern.After that,DFU-Net introduces CLC-UNet module on the basis of U-Net to extract richer features and context information,and The DFU-Net model’s processing of image segmentation tasks resolves the issue of spatial information being lost.The comparison model is outdone by this model in terms of segmentation performance,as evidenced by the experimental results.(3)Develop a deep learning intelligent diagnosis system for dividing and merging pathological images of breast cancer.On the basis of the research on pathological image classification based on SE-Bi Pool Net and pathological image segmentation algorithm based on DFU Net,an in-depth learning intelligent diagnosis system for dividing and merging breast cancer frozen images is developed,and test cases are written for system testing.