Blockchain Digital Twin for Stroke Prediction

Blockchain-enabled digital twin system for brain stroke prediction

• A digital twin is a virtual model of a real-world system, updating in real-time
• Digital twins are gaining popularity in healthcare for monitoring activities like diet, physical activity, and sleep
• Current research shows limited accuracy in predicting serious conditions like heart attacks, brain strokes, and cancers using digital twins
• Data security and privacy issues continue to hinder widespread adoption of digital twin models
• A secure, machine learning-powered digital twin application was developed to enhance prediction accuracy, strengthen security, and ensure scalability
• The application demonstrated 98.28% accuracy for brain stroke prediction on a selected dataset
• Consortium blockchain technology was integrated with machine learning for enhanced data security, creating a tamper-proof system
• Data anomalies are automatically corrected to maintain robust data protection
• The application can be adapted to monitor other pathologies like heart attacks, cancers, osteoporosis, and epilepsy with minimal adjustments

Introduction

• A digital twin is a digital representation of a physical entity (e.g., person or machine)
• The industrial and academic research communities have significantly advanced the use of digital twins in various applications, including industrial processes, aerospace, and healthcare
• In healthcare, digital twins are utilized in hospitals to analyze operational changes, evaluate staffing levels, and optimize care delivery
• Remote patient monitoring leverages the benefits of digital twins

Literature Review

• Digital twin frameworks, which virtually replicate physical systems using patient data and advanced computational techniques, have shown promise in predicting pathologies like brain stroke
• Several studies have explored the application of digital twin technology, striving to enhance stroke prediction accuracy and improve patient outcomes
• Key studies examined in the literature review include those proposing digital twin frameworks that leverage machine learning algorithms for prediction accuracy (accuracy levels varied)
• Concerns have been voiced regarding the lack of security and privacy measures in several prior studies.

Methods

• A machine learning (ML) based Digital Twin application for predicting brain strokes was developed
• The application leverages a public dataset from Kaggle (4,981 records)
• The dataset includes patient data, such as age, sex, existing medical history, and lifestyle factors
• Three files (Patient EHR, supplementary data, and transactional data) were created to simulate patient data
• The data in these files was simulated, but drew some information from the Kaggle dataset
• Data was stored in XML format, for integration with wearable devices
• Data from wearables and other sources is stored in a CSV file
• Medical device information, combined with clinical history, is stored on a blockchain

Results

• A digital twin application was developed with an accuracy exceeding 98.28% for stroke prediction
• The application was designed to adapt to other diseases (heart attacks, cancers, osteoporosis, and epilepsy) with minimal code changes
• The system successfully detected deliberate data tampering, demonstrating the security benefits of using blockchain in healthcare

Discussion

• Security and privacy are crucial concerns in healthcare, potentially impacting the implementation of digital twin solutions
• This model leverages a consortium blockchain for robust data security
• The application’s scalability makes it adaptable to various critical illnesses with minimal code adjustments
• The model's accuracy (98.28%) exceeds those in previous studies (generally ranging from 84–92%)

Conclusion

• A digital twin application was successfully developed to predict stroke
• The model demonstrates capabilities for prediction, scalability, and security
• Further research is proposed to evaluate the model across more diverse and real-world datasets
• Adaptability for other pathologies is demonstrated