Digital image analysis in breast pathology—from image processing techniques to artificial intelligence

Robertson, Stephanie; Azizpour, Hossein; Smith, Kevin; Hartman, Johan

doi:10.1016/j.trsl.2017.10.010

Cited by 238 publications

(154 citation statements)

References 163 publications

(240 reference statements)

Supporting

Mentioning

134

Contrasting

Unclassified

Order By: Relevance

“…In this context, the adoption of digital pathology and software‐assisted methods may increase accuracy, reduce human error, and ultimately improve reproducibility of PD‐L1 assessment and interpretation . Recently, PD‐L1 expression measured by IHC and assessed by digital pathology platforms has been positively associated with outcome in two cohorts of patients with TN early BC treated with surgery and standard CT .…”

Section: Methodsmentioning

confidence: 99%

Programmed Cell Death Ligand 1 in Breast Cancer: Technical Aspects, Prognostic Implications, and Predictive Value

2019

View full text Add to dashboard Cite

In the light of recent advances in the immunotherapy field for breast cancer (BC) treatment, especially in the triple‐negative subtype, the identification of reliable biomarkers capable of improving patient selection is paramount, because only a portion of patients seem to derive benefit from this appealing treatment strategy. In this context, the role of programmed cell death ligand 1 (PD‐L1) as a potential prognostic and/or predictive biomarker has been intensively explored, with controversial results. The aim of the present review is to collect available evidence on the biological relevance and clinical utility of PD‐L1 expression in BC, with particular emphasis on technical aspects, prognostic implications, and predictive value of this promising biomarker. Implications for Practice In the light of the promising results coming from trials of immune checkpoint inhibitors for breast cancer treatment, the potential predictive and/or prognostic role of programmed cell death ligand 1 (PD‐L1) in breast cancer has gained increasing interest. This review provides clinicians with an overview of the available clinical evidence regarding PD‐L1 as a biomarker in breast cancer, focusing on both data with a possible direct impact on clinic and methodological pitfalls that need to be addressed in order to optimize PD‐L1 implementation as a clinically useful tool for breast cancer management.

show abstract

Section: Methodsmentioning

confidence: 99%

Programmed Cell Death Ligand 1 in Breast Cancer: Technical Aspects, Prognostic Implications, and Predictive Value

2019

View full text Add to dashboard Cite

show abstract

“…Breast carcinomas arise from the mammary epithelium and cause a pre-malignant epithelial proliferation within the ducts, called ductal carcinoma in situ. Invasive carcinoma is characterized by the cancer cells gaining the capacity to break through the basal membrane of the duct walls and infiltrate into surrounding tissues [21].…”

Section: Introductionmentioning

confidence: 99%

Deep Convolutional Neural Networks for Breast Cancer Histology Image Analysis

Rakhlin

Shvets

Iglovikov³

et al. 2018

Preprint

114

126

View full text Add to dashboard Cite

Abstract. Breast cancer is one of the main causes of cancer death worldwide. Early diagnostics significantly increases the chances of correct treatment and survival, but this process is tedious and often leads to a disagreement between pathologists. Computer-aided diagnosis systems showed potential for improving the diagnostic accuracy. In this work, we develop the computational approach based on deep convolution neural networks for breast cancer histology image classification. Hematoxylin and eosin stained breast histology microscopy image dataset is provided as a part of the ICIAR Grand Challenge on Breast Cancer Histology Images. Our approach utilizes several deep neural network architectures and gradient boosted trees classifier. For -class classification task, we report . % accuracy. For -class classification task to detect carcinomas we report . % accuracy, AUC . %, and sensitivity/specificity . / . % at the high-sensitivity operating point. To our knowledge, this approach outperforms other common methods in automated histopathological image classification. The source code for our approach is made publicly available at https://github.com/alexander-rakhlin/ICIAR

show abstract

“…19 Many agree that transfer learning techniques provide the most useful path forwards in CNN image analysis. 3,[20][21][22] However, there are few reports in the pathology literature that attempt to explain which pre-analytical variables should be controlled when ML/AI models are created to evaluate The difference in mean accuracy between PNG and JPG models for colon ResNet50 with PNGs was 80.6%, and the mean accuracy with JPGs was 82.3%. This represents a statistically significant but clinically insubstantial difference.…”

Section: Discussionmentioning

confidence: 99%

Impact of pre‐analytical variables on deep learning accuracy in histopathology

et al. 2019

View full text Add to dashboard Cite

Aims Machine learning (ML) binary classification in diagnostic histopathology is an area of intense investigation. Several assumptions, including training image quality/format and the number of training images required, appear to be similar in many studies irrespective of the paucity of supporting evidence. We empirically compared training image file type, training set size, and two common convolutional neural networks (CNNs) using transfer learning (ResNet50 and SqueezeNet). Methods and results Thirty haematoxylin and eosin (H&E)‐stained slides with carcinoma or normal tissue from three tissue types (breast, colon, and prostate) were photographed, generating 3000 partially overlapping images (1000 per tissue type). These lossless Portable Networks Graphics (PNGs) images were converted to lossy Joint Photographic Experts Group (JPG) images. Tissue type‐specific binary classification ML models were developed by the use of all PNG or JPG images, and repeated with a subset of 500, 200, 100, 50, 30 and 10 images. Eleven models were generated for each tissue type, at each quantity of training images, for each file type, and for each CNN, resulting in 924 models. Internal accuracies and generalisation accuracies were compared. There was no meaningful significant difference in accuracies between PNG and JPG models. Models trained with more images did not invariably perform better. ResNet50 typically outperformed SqueezeNet. Models were generalisable within a tissue type but not across tissue types. Conclusions Lossy JPG images were not inferior to lossless PNG images in our models. Large numbers of unique H&E‐stained slides were not required for training optimal ML models. This reinforces the need for an evidence‐based approach to best practices for histopathological ML.

show abstract

Digital image analysis in breast pathology—from image processing techniques to artificial intelligence

Cited by 238 publications

References 163 publications

Programmed Cell Death Ligand 1 in Breast Cancer: Technical Aspects, Prognostic Implications, and Predictive Value

Programmed Cell Death Ligand 1 in Breast Cancer: Technical Aspects, Prognostic Implications, and Predictive Value

Deep Convolutional Neural Networks for Breast Cancer Histology Image Analysis

Impact of pre‐analytical variables on deep learning accuracy in histopathology

Contact Info

Product

Resources

About