Federated learning for computational pathology on gigapixel whole slide images

Lu, Ming; Chen, Richard J.; Kong, Dehan; Lipková, Jana; Singh, Vikram; Williamson, Drew F. K.; Chen, Tiffany; Mahmood, Faisal

doi:10.1016/j.media.2021.102298

Cited by 137 publications

(91 citation statements)

References 55 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Data sharing between institutions may require patients to forfeit their rights of data control. This problem has been tackled by (centralized) federated learning (FL) 23 , 24 , in which multiple AI models are trained independently on separate computers (peers). In FL, peers do not share any input data with each other, and only share the learned model weights.…”

Section: Mainmentioning

confidence: 99%

Swarm learning for decentralized artificial intelligence in cancer histopathology

Saldanha

Quirke

West

et al. 2022

Nat Med

121

View full text Add to dashboard Cite

Artificial intelligence (AI) can predict the presence of molecular alterations directly from routine histopathology slides. However, training robust AI systems requires large datasets for which data collection faces practical, ethical and legal obstacles. These obstacles could be overcome with swarm learning (SL), in which partners jointly train AI models while avoiding data transfer and monopolistic data governance. Here, we demonstrate the successful use of SL in large, multicentric datasets of gigapixel histopathology images from over 5,000 patients. We show that AI models trained using SL can predict BRAF mutational status and microsatellite instability directly from hematoxylin and eosin (H&E)-stained pathology slides of colorectal cancer. We trained AI models on three patient cohorts from Northern Ireland, Germany and the United States, and validated the prediction performance in two independent datasets from the United Kingdom. Our data show that SL-trained AI models outperform most locally trained models, and perform on par with models that are trained on the merged datasets. In addition, we show that SL-based AI models are data efficient. In the future, SL can be used to train distributed AI models for any histopathology image analysis task, eliminating the need for data transfer.

show abstract

Section: Mainmentioning

confidence: 99%

Swarm learning for decentralized artificial intelligence in cancer histopathology

Saldanha

Quirke

West

et al. 2022

Nat Med

121

View full text Add to dashboard Cite

show abstract

“…On one side, whole-slide images are large: making them available requires data storage and bandwidth, which come at a cost. The images are also typically saved in bespoke, scanner-specific file formats that include additional labels and metadata; this makes them difficult to anonymise both fully and with confidence that nothing has been missed [62]. Sharing also raises important questions around how the images may be used, including whether it is permitted to develop commercial AI models based on them; these must be resolved, with clear license statements provided, for others to safely use the shared data [63,64].…”

Section: Complications Of Sharingmentioning

confidence: 99%

Developing image analysis methods for digital pathology

Bankhead

2022

The Journal of Pathology

View full text Add to dashboard Cite

The potential to use quantitative image analysis and artificial intelligence is one of the driving forces behind digital pathology. However, despite novel image analysis methods for pathology being described across many publications, few become widely adopted and many are not applied in more than a single study. The explanation is often straightforward: software implementing the method is simply not available, or is too complex, incomplete, or dataset‐dependent for others to use. The result is a disconnect between what seems already possible in digital pathology based upon the literature, and what actually is possible for anyone wishing to apply it using currently available software. This review begins by introducing the main approaches and techniques involved in analysing pathology images. I then examine the practical challenges inherent in taking algorithms beyond proof‐of‐concept, from both a user and developer perspective. I describe the need for a collaborative and multidisciplinary approach to developing and validating meaningful new algorithms, and argue that openness, implementation, and usability deserve more attention among digital pathology researchers. The review ends with a discussion about how digital pathology could benefit from interacting with and learning from the wider bioimage analysis community, particularly with regard to sharing data, software, and ideas. © 2022 The Author. The Journal of Pathology published by John Wiley & Sons Ltd on behalf of The Pathological Society of Great Britain and Ireland.

show abstract

“…A wide range of applications demonstrate that federated learning is a potential fit for leveraging diverse types of medical data, including electronic health records [16], genomic data [17], and time-series data from wearables [18]. Examples related to medical image classification include brain tumor segmentation [9,19], classification and survival prediction on whole slide images in pathology [20], classification of functional magnetic resonance images (fMRI) [21], and breast density classification from mammographic images [22]. One large research area is concerned with the classification of chest X-ray images.…”

Section: Related Workmentioning

confidence: 99%

“…Differential privacy in federated learning is often achieved using differentially-private stochastic gradient descent (DP-SGD) [7,41,42], an algorithm that determines the appropriate noise scale and how to clip the model parameter. The combination of federated learning and differential privacy has been explored in multiple medical use cases including prediction of mortality and adverse drug reactions from electronic health records [43], brain tumor segmentation [9], classification of pathology whole slide images [20], detection of diabetic retinopathy in images of the retina [44], and identification of lung cancer in histopathologic images [45].…”

Section: Related Workmentioning

confidence: 99%

Defending against Reconstruction Attacks through Differentially Private Federated Learning for Classification of Heterogeneous Chest X-Ray Data

Ziegler¹,

Pfitzner²,

Schulz³

et al. 2022

Preprint

View full text Add to dashboard Cite

Privacy regulations and the physical distribution of heterogeneous data are often primary concerns for the development of deep learning models in a medical context. This paper evaluates the feasibility of differentially private federated learning for chest X-ray classification as a defense against data privacy attacks. To the best of our knowledge, we are the first to directly compare the impact of differentially private training on two different neural network architectures, DenseNet121 and ResNet50. Extending the federated learning environments previously analyzed in terms of privacy, we simulated a heterogeneous and imbalanced federated setting by distributing images from the public CheXpert and Mendeley chest X-ray datasets unevenly among 36 clients. Both non-private baseline models achieved an area under the receiver operating characteristic curve (AUC) of 0.94 on the binary classification task of detecting the presence of a medical finding. We demonstrate that both model architectures are vulnerable to privacy violation by applying image reconstruction attacks to local model updates from individual clients. The attack was particularly successful during later training stages. To mitigate the risk of privacy breach, we integrated Rényi differential privacy with a Gaussian noise mechanism into local model training. We evaluate model performance and attack vulnerability for privacy budgets ε ∈ {1, 3, 6, 10}. The DenseNet121 achieved the best utility-privacy trade-off with an AUC of 0.94 for ε = 6. Model performance deteriorated slightly for individual clients compared to the non-private baseline. The ResNet50 only reached an AUC of 0.76 in the same privacy setting. Its performance was inferior to that of the DenseNet121 for all considered privacy constraints, suggesting that the DenseNet121 architecture is more robust to differentially private training.

show abstract

Federated learning for computational pathology on gigapixel whole slide images

Cited by 137 publications

References 55 publications

Swarm learning for decentralized artificial intelligence in cancer histopathology

Swarm learning for decentralized artificial intelligence in cancer histopathology

Developing image analysis methods for digital pathology

Defending against Reconstruction Attacks through Differentially Private Federated Learning for Classification of Heterogeneous Chest X-Ray Data

Contact Info

Product

Resources

About