[Paper Review] Blockchain-Federated-Learning and Deep Learning Models for COVID-19 Detection Using CT Imaging

Rajesh Kumar, Abdullah Aman Khan+8 more|arXiv (Cornell University)|Jul 10, 2020

COVID-19 diagnosis using AI72 references466 citations

TL;DR

This paper proposes a blockchain-federated learning framework that enables privacy-preserving collaborative training of a Capsule Network for COVID-19 detection from CT scans. By normalizing heterogeneous data from multiple hospitals and sharing only model updates via a blockchain-secured federated learning pipeline, the method achieves 98.68% accuracy in classifying COVID-19, viral pneumonia, and normal cases, while preserving data privacy and enabling real-time model improvement across institutions.

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.

Motivation & Objective

• To address the challenge of sharing sensitive medical data across hospitals while preserving patient and institutional privacy.
• To develop a robust deep learning model capable of accurately detecting early signs of COVID-19 in lung CT scans despite data heterogeneity from diverse CT scanners.
• To enable collaborative global model training without centralizing raw data, overcoming data silos and privacy barriers in healthcare AI.
• To introduce a new, publicly available CT scan dataset (CC-19) comprising 34,006 slices from 89 patients, including 68 confirmed COVID-19 cases.

Proposed method

• Proposed a data normalization technique to handle modality and scanner heterogeneity across multi-source CT imaging data.
• Employed a Capsule Network for both segmentation and classification of lung abnormalities to improve generalization and feature learning.
• Integrated federated learning to train a global deep learning model by aggregating local model updates from hospitals without sharing raw data.
• Utilized blockchain technology to authenticate and securely log model updates and data contributions, ensuring transparency and integrity in the collaborative training process.
• Designed a decentralized, privacy-preserving framework where hospitals retain local data and only exchange model weights and gradients.
• Released a new public dataset named CC-19, containing 34,006 CT slices from 89 subjects (68 positive, 21 negative) for open research use.

Experimental results

Research questions

• RQ1 Can a federated learning framework with blockchain integration effectively train a global deep learning model for COVID-19 detection without compromising data privacy across hospitals?
• RQ2 How does data heterogeneity from different CT scanners affect model performance, and can normalization techniques mitigate this issue?
• RQ3 Does a Capsule Network outperform conventional CNNs in classifying subtle COVID-19 patterns in lung CT scans?
• RQ4 To what extent can collaborative training via federated learning improve diagnostic accuracy compared to isolated model training on single-hospital data?
• RQ5 Can a transparent, immutable blockchain ledger enhance trust and accountability in multi-institutional medical AI model development?

Key findings

• The proposed framework achieved a classification accuracy of 98.68% on the CC-19 dataset, significantly outperforming existing methods.
• The Capsule Network-based model demonstrated superior generalization and sensitivity (98%) compared to traditional CNNs, particularly in detecting small, early-stage infection regions.
• Federated learning enabled effective global model training across three hospitals, with model accuracy converging steadily over 80 training iterations.
• The training loss decreased consistently across all hospitals, indicating stable and effective model convergence in the decentralized setup.
• The blockchain-based system ensured secure, transparent, and auditable model updates, with no data leakage observed during the collaborative training process.
• The public release of the CC-19 dataset provides a valuable benchmark for future research in AI-driven COVID-19 detection using medical imaging.