Blockchain-Federated-Learning and Deep Learning Models for COVID-19 detection using CT Imaging (2007.06537v2)

Abstract: With the increase of COVID-19 cases worldwide, an effective way is required to diagnose COVID-19 patients. The primary problem in diagnosing COVID-19 patients is the shortage and reliability of testing kits, due to the quick spread of the virus, medical practitioners are facing difficulty identifying the positive cases. The second real-world problem is to share the data among the hospitals globally while keeping in view the privacy concerns of the organizations. Building a collaborative model and preserving privacy are major concerns for training a global deep learning model. This paper proposes a framework that collects a small amount of data from different sources (various hospitals) and trains a global deep learning model using blockchain based federated learning. Blockchain technology authenticates the data and federated learning trains the model globally while preserving the privacy of the organization. First, we propose a data normalization technique that deals with the heterogeneity of data as the data is gathered from different hospitals having different kinds of CT scanners. Secondly, we use Capsule Network-based segmentation and classification to detect COVID-19 patients. Thirdly, we design a method that can collaboratively train a global model using blockchain technology with federated learning while preserving privacy. Additionally, we collected real-life COVID-19 patients data, which is, open to the research community. The proposed framework can utilize up-to-date data which improves the recognition of computed tomography (CT) images. Finally, our results demonstrate a better performance to detect COVID-19 patients.

• The paper presents a novel framework combining blockchain, federated learning, and a Capsule Network to securely detect COVID-19 from CT scans.
• The methodology standardizes heterogeneous CT data and enables decentralized model updates, resulting in superior sensitivity and specificity.
• The study introduces the CC-19 dataset, fostering collaborative research and future innovations in privacy-preserving medical AI.

Blockchain-Federated Learning and Deep Learning Models for COVID-19 Detection using CT Imaging

The paper at hand presents a detailed framework integrating blockchain technology with federated learning to address the challenges of privacy-preserved data sharing and the global training of a deep learning model for the detection of COVID-19 from CT images. This work is positioned at the intersection of machine learning, blockchain, and medical imaging, aiming to collaboratively enhance the diagnostic capability without infringing on patient privacy.

The authors begin by contextualizing the necessity of such a paper in light of the COVID-19 pandemic, noting the significant strain on available testing kits and the critical need for reliable diagnostic solutions. They also emphasize the complications involved in data sharing among hospitals due to privacy concerns. Traditional deep learning methods are hamstrung by data availability limitations, and this framework seeks to mitigate such issues through an innovative application of blockchain and federated learning.

Methodology Overview

The authors articulate a multi-step approach:

1. Data Normalization: The paper proposes a data normalization technique to address data heterogeneity, crucial given the diverse CT scanning equipment used across different healthcare facilities. This involves spatial and signal normalization to align the CT image data to a common standard.
2. Deep Learning Model: A Capsule Network-based architecture is employed for both the segmentation and classification of CT images, aiming to efficiently detect COVID-19. The Capsule Network reportedly offers superior performance, enhancing the feature extraction process by replacing traditional max-pooling operations with routing-by-agreement mechanisms.
3. Federated Learning Implementation: The framework leverages federated learning to combine machine learning models trained at different hospitals, therefore allowing the global model to benefit from broader data diversity without compromising on data privacy.
4. Blockchain Integration: Utilization of blockchain technology ensures a secure means of authenticating, distributing, and storing model weights and gradients among participating hospitals. This decentralized approach relieves participant hospitals of privacy concerns, only exchanging learned model parameters instead of raw data.

Results and Implications

A notable contribution of the paper is the introduction of a novel dataset, named CC-19, containing over 34,000 CT scan slices which have been made public for further research. Experimental evaluations demonstrate that the proposed model, using the Capsule Network, outperforms established deep learning models across multiple performance metrics, particularly in sensitivity and specificity—a critical requirement for medical diagnostics.

The integration of blockchain offers a security advantage, ensuring verifiable execution of federated learning while simultaneously providing a mechanism for maintaining data integrity across decentralized nodes. This combination showcases a viable pathway for real-world deployment, hinting at potential extensibility to other medical conditions and imaging modalities.

Conclusion and Future Directions

The authors conclude with reflections on the potential application of their framework as a predictive tool in healthcare systems worldwide. This approach exemplifies a feasible solution to the twin challenges of collaborative model training and privacy preservation, drawing on advanced computational techniques. The success in operationalizing federated learning through blockchain could open avenues for other areas of AI research where privacy is paramount.

Future research may expand on the scalability of such a system, exploring how different optimal configurations could apply to various datasets and medical challenges beyond COVID-19. Additionally, reducing the computational overhead inherent in blockchain operations, without compromising data security, will be an area for innovation.

Through the confluence of AI, blockchain, and federated learning, this paper points to an emergent suite of technologies poised to redefine collaborative efforts in biomedical research and beyond.