Learning Domain-Agnostic Visual Representation for Computational Pathology Using Medically-Irrelevant Style Transfer Augmentation

@article{Yamashita2021LearningDV,
  title={Learning Domain-Agnostic Visual Representation for Computational Pathology Using Medically-Irrelevant Style Transfer Augmentation},
  author={Rikiya Yamashita and Jin Long and Snikitha Banda and Jeanne Shen and D. Rubin},
  journal={IEEE Transactions on Medical Imaging},
  year={2021},
  volume={40},
  pages={3945-3954}
}
Suboptimal generalization of machine learning models on unseen data is a key challenge which hampers the clinical applicability of such models to medical imaging. Although various methods such as domain adaptation and domain generalization have evolved to combat this challenge, learning robust and generalizable representations is core to medical image understanding, and continues to be a problem. Here, we propose STRAP (Style TRansfer Augmentation for histoPathology), a form of data… 
How GNNs Facilitate CNNs in Mining Geometric Information from Large-Scale Medical Images
TLDR
A fusion framework for enhancing the global image-level representation captured by CNNs with the geometry of cell-level spatial information learned by graph neural networks (GNN), which optimizes an integration between collaborative features of global images and cell graphs.
Benchmarking the Robustness of Deep Neural Networks to Common Corruptions in Digital Pathology
TLDR
An easy-to-use benchmark is established to evaluate how deep neural networks perform on corrupted pathology images to guarantee the robustness of the model with respect to a wide range of image corruptions.
Test Time Transform Prediction for Open Set Histopathological Image Recognition
TLDR
A new approach for Open Set histopathological image recognition is introduced based on training a model to accurately identify image categories and simultaneously predict which data augmentation transform has been applied, which is expected to be lower for images in the Open Set.
MaxStyle: Adversarial Style Composition for Robust Medical Image Segmentation
TLDR
A novel data augmentation framework called MaxStyle is proposed, which maximizes the effectiveness of style augmentation for model OOD performance and attaches an auxiliary style-augmented image decoder to a segmentation network for robust feature learning andData augmentation.
Built to last? Reproducibility and Reusability of Deep Learning Algorithms in Computational Pathology
TLDR
This review highlights candidates for reproducible and reusable algorithms in computational pathology, and provides a list of reusable data handling tools and a detailed overview of the publications together with the criteria for reproducibility and reusability.
A Histopathology Study Comparing Contrastive Semi-Supervised and Fully Supervised Learning
TLDR
For the task explored in this paper, it is found that ImageNet pre-trained networks largely outperform the self-supervised representations obtained using Barlow Twins.
Weakly supervised annotation‐free cancer detection and prediction of genotype in routine histopathology
TLDR
The Slide‐Level Assessment Model (SLAM) uses a single off‐the‐shelf neural network to predict molecular alterations directly from routine pathology slides without any manual annotations, improving upon previous methods by automatically excluding normal and non‐informative tissue regions.
Making Radiomics More Reproducible across Scanner and Imaging Protocol Variations: A Review of Harmonization Methods
TLDR
A narrative review of various harmonization solutions to make the radiomic features more reproducible across various scanners and protocol settings especially using generative adversarial networks (GANs), neural style transfer (NST) techniques, or a combination of both.
...
...

References

SHOWING 1-10 OF 37 REFERENCES
Histological images for MSI vs. MSS classification in gastrointestinal cancer, FFPE samples [data set] Zenodo
  • 2019
WILDS: A Benchmark of in-the-Wild Distribution Shifts
TLDR
WILDS is presented, a benchmark of in-the-wild distribution shifts spanning diverse data modalities and applications, and is hoped to encourage the development of general-purpose methods that are anchored to real-world distribution shifts and that work well across different applications and problem settings.
StyPath: Style-Transfer Data Augmentation For Robust Histology Image Classification
TLDR
A novel pipeline to build robust deep neural networks for AMR classification based on StyPath, a histological data augmentation technique that leverages a light weight style-transfer algorithm as a means to reduce sample-specific bias and generalization ability is proposed.
Clinical-grade Detection of Microsatellite Instability in Colorectal Tumors by Deep Learning.
TLDR
A deep-learning system that detects colorectal cancer specimens with dMMR or MSI using hematoxylin and eosin-stained slides is developed and might be used for high-throughput, low-cost evaluation of coloreCTal tissue specimens.
Learning to Learn Single Domain Generalization
TLDR
A new method named adversarial domain augmentation is proposed to solve the Out-of-Distribution (OOD) generalization problem by leveraging adversarial training to create "fictitious" yet "challenging" populations, from which a model can learn to generalize with theoretical guarantees.
A Simple Framework for Contrastive Learning of Visual Representations
TLDR
It is shown that composition of data augmentations plays a critical role in defining effective predictive tasks, and introducing a learnable nonlinear transformation between the representation and the contrastive loss substantially improves the quality of the learned representations, and contrastive learning benefits from larger batch sizes and more training steps compared to supervised learning.
Momentum Contrast for Unsupervised Visual Representation Learning
We present Momentum Contrast (MoCo) for unsupervised visual representation learning. From a perspective on contrastive learning as dictionary look-up, we build a dynamic dictionary with a queue and a
Key challenges for delivering clinical impact with artificial intelligence
TLDR
The safe and timely translation of AI research into clinically validated and appropriately regulated systems that can benefit everyone is challenging, and robust clinical evaluation, using metrics that are intuitive to clinicians and ideally go beyond measures of technical accuracy, is essential.
...
...