數(shù)字病理學(xué)是醫(yī)學(xué)協(xié)議中具有挑戰(zhàn)性的進(jìn)展之一。病理檢查在診斷過程中起著至關(guān)重要的作用,并使病理學(xué)家能夠?qū)ξ⒂^結(jié)構(gòu)進(jìn)行分類。病理學(xué)家在顯微鏡下分析了大量的活檢切片。對(duì)細(xì)胞核的組織學(xué)結(jié)構(gòu),形態(tài)變化和生物組織分布的分析有助于病理學(xué)家更好地識(shí)別組織病理學(xué)樣本。高含量活檢組織病理學(xué)分類和分級(jí)可提供重要的預(yù)后信息,這對(duì)于了解疾�。ò┌Y）的擴(kuò)散和預(yù)報(bào)至關(guān)重要。在主要的癌癥中,乳腺癌是影響世界各地女性癌癥死亡的主要原因之一,而且在發(fā)達(dá)國家和發(fā)展中國家逐漸增加。然而,在乳腺組織的全切片掃描圖像（Whole slide images,WSIs）中手動(dòng)診斷疾病是一項(xiàng)艱巨而富有挑戰(zhàn)性的任務(wù)。為了克服手動(dòng)分析的缺點(diǎn),各種自動(dòng)診斷系統(tǒng)（即,簡單的圖像處理算法或基于深度學(xué)習(xí)的算法）相繼被開發(fā)了出來。本文主要研究開發(fā)基于深度學(xué)習(xí)的自動(dòng)檢測(cè)框架,這個(gè)框架能夠從不同類型的組織病理學(xué)圖像中對(duì)乳腺癌進(jìn)行檢測(cè)。世界衛(wèi)生組織提出了一種乳腺癌分級(jí)標(biāo)準(zhǔn),稱為諾丁漢分級(jí)系統(tǒng)。它結(jié)合了三種形態(tài)學(xué)預(yù)后因素,即有絲分裂計(jì)數(shù),腎小管形成和細(xì)胞核多形性。本文描述了有關(guān)組織病理學(xué)圖像分析的理論指導(dǎo),并定義了可提高組織病理學(xué)家診斷和預(yù)報(bào)能力的乳腺癌分... 

【文章來源】：哈爾濱工業(yè)大學(xué)黑龍江省 211工程院校 985工程院校

【文章頁數(shù)】：156 頁

【學(xué)位級(jí)別】：博士

【文章目錄】：
摘要
Abstract
Acronyms and symbols
Chapter 1 Introduction
    1.1 Background and significances of the study
    1.2 Basic theory of deep learning
        1.2.1 Deep learning
        1.2.2 Deep learning programming frameworks
    1.3 Steps involved in biopsy slide preparation and breast cancer grading
        1.3.1 Fast slide scanners for digital image acquisition
        1.3.2 Standard grading system for breast cancer recognition
    1.4 Research status of Histopathology image detection techniques
        1.4.1 Breast cancer detection/localization methods
        1.4.2 Breast cancer segmentation methods
        1.4.3 Breast cancer classification methods
    1.5 Motivation
    1.6 Limitations and challenges in breast histology image analysis
    1.7 Thesis structure
    1.8 Contributions
Chapter 2 Stain color normalization of hematoxylin and eosin stained histopathologyimages
    2.1 Introduction
    2.2 Stain color normalization of hematoxylin and eosin stained images
        2.2.1 Benchmark color normalization methods to reduce color variation in breasthistology images
        2.2.2 Proposed color normalization method based on image structure and color statics
    2.3 Experimental results
        2.3.1 Materials
        2.3.2 Metrics to check the performance of color normalization methods
        2.3.3 Experimental results and performance analysis
    2.4 Summary
Chapter 3 Mitosis detection in breast histology with fully fused and multi-scale fully fusedconvolution neural networks
    3.1 Introduction
    3.2 Pre-processing of the Hematoxylin & Eosin stained microscopy images
        3.2.1 Stain-normalization of Hematoxylin & Eosin stained images
        3.2.2 Patch extraction （coarsely extracted patches） from high resolution images
        3.2.3 Patch extraction （fine discriminative patches） using sample selection strategy
    3.3 Deep learning based methods
        3.3.1 Features fused convolution neural network for mitosis detection
        3.3.2 Multi-scale feature fused convolution neural network for mitosis detection
    3.4 Experiments and results
        3.4.1 Candidate detection on MITOS-ATYPIA-14 validation dataset using varioustypes of extracted patches
        3.4.2 Visual results evaluated with FF-CNN and MFF-CNN models on validation HPFimages
        3.4.3 Detection performance of FF-CNN and MFF-CNN models on test dataset
    3.5 Summary
Chapter 4 Small mitotic cells detection using multi-scale object detector and atrous fullyconvolution based deep segmentation model
    4.1 Introduction
    4.2 Contributions
    4.3 Publically available Mitosis datasets
    4.4 Deep learning based Mitosis detection frameworks
        4.4.1 Multi-scale region proposal model for mitotic cells detection
        4.4.2 Atrous fully convolution model for bounding box estimation
    4.5 Experimental analysis and results
        4.5.1 Performance on ICPR 2012 grand challenge mitosis dataset
        4.5.2 Performance on ICPR 2014 and AMIDA13 grand challenge mitosis dataset
    4.6 Summary
Chapter 5 Multi-class breast cancer recognition with wavelet decomposed image basedconvolution neural network
    5.1 Introduction
    5.2 Contributions
    5.3 Pre-processing of H&E stained histopathological images
        5.3.1 Multi-class breast cancer histology datasets
        5.3.2 Color normalization of breast histology images
        5.3.3 Data augmentation with rotation technique
        5.3.4 Data augmentation with proposed channel color augmentations algorithm
        5.3.5 Theory of Wavelet transform
        5.3.6 Haar wavelet decomposition of the breast histology images
    5.4 Proposed deep convolution neural network based model for multi-class classification
    5.5 Image classification techniques
        5.5.1 Classification with Softmax classifier
        5.5.2 Classification with support vector machine , k-nearest neighbor and Randomforest classifiers
    5.6 Computational complexity of CNN models
        5.6.1 Computations reduction of proposed HWDCNN models with Haar waveletdecomposed images
    5.7 Experimental analysis and results
        5.7.1 Performance on ICIAR 2018 multi-class dataset
        5.7.2 Performance on Break His multi class cancer dataset
    5.8 Summary
Conclusion
References
List of publications
Acknowledgements
Resume Details



本文編號(hào)：3007277