Blockchain-Federated and Deep-Learning-Based Ensembling of Capsule Network with Incremental Extreme Learning Machines...

Area of Science:

• Artificial Intelligence in Medical Imaging
• Deep Learning for Disease Detection
• Blockchain in Healthcare

Background:

• Rapid SARS-CoV-2 dissemination necessitates effective COVID-19 isolation strategies.
• Challenges in COVID-19 diagnosis include rapid virus propagation and lack of reliable testing models.
• Sharing medical data between institutions is hindered by privacy concerns, complicating global model training.

Purpose of the Study:

• To develop a novel framework for training a global deep learning model for COVID-19 detection.
• To address challenges in data sharing and user privacy in collaborative model development.
• To improve the accuracy and efficiency of COVID-19 diagnosis using medical imaging.

Main Methods:

• A blockchain-based federated learning (FL) framework was designed to compile data from five hospital databases.
• Data normalization techniques were employed to handle diversity from multiple CT scanners.
• A hybrid model combining Capsule Network (CapsNet) and Incremental Extreme Learning Machines (IELMs) was used for patient categorization.
• Blockchain Technology (BCT) was utilized for data validation and ensuring anonymity during interactive model training.

Main Results:

• The proposed framework achieved a high accuracy of 98.99% in identifying COVID-19 patients using chest CT scans.
• The model demonstrated superior classification performance compared to five other deep learning algorithms.
• The framework successfully maintained data privacy throughout the collaborative training process.

Conclusions:

• The developed framework offers a robust solution for global deep learning model training in healthcare.
• Blockchain-based federated learning effectively addresses privacy concerns in collaborative medical data analysis.
• The model significantly aids medical practitioners in the accurate and private diagnosis of COVID-19.