An Evaluation of Continuous Integration Strategies for Deploying Machine Learning Models in Healthcare Applications
Keywords:
Continuous Integration (CI), Machine Learning (ML), Healthcare AI, Deployment Pipelines, MLOps, Compliance, ReproducibilityAbstract
The deployment of machine learning (ML) models in healthcare presents unique challenges due to stringent regulatory standards, data sensitivity, and the need for real-time performance. Continuous Integration (CI) strategies offer a potential solution for streamlining ML model deployment, ensuring model robustness, and enabling rapid iterations. This paper evaluates multiple CI strategies adapted for ML applications in healthcare, focusing on model validation, reproducibility, versioning, and compliance. Through a comparative analysis of existing approaches, we highlight the advantages and limitations of current CI pipelines and propose an optimized flow for healthcare-specific needs. Empirical results from selected healthcare ML projects underscore the performance and compliance benefits of robust CI practices. A line graph illustrates the reduction in deployment errors with increasing CI pipeline maturity across selected deployments.
References
Zhang, Li, and Min Xu. "An End-to-End Evaluation of ML Pipelines in Radiology." IEEE Transactions on Medical Imaging, vol. 41, no. 3, 2022.
Kapoor, Sanjay, et al. "Explainability in CI Pipelines for Cardiological ML Models." Journal of Biomedical Informatics, vol. 132, no. 4, 2023.
Kavuri, S. (2022). Large Language Model (LLM)-Based Automation for Software Test Script Generation. Computer Fraud & Security, 2022(11), 17–28.
Chen, Alex, and Sunita Pal. "Automated Testing Frameworks for Compliant ML Models." ACM HealthTech Journal, vol. 24, no. 1, 2021.
Anand, Ritu, et al. "Drift Monitoring in Clinical Sepsis Prediction Systems." Nature Digital Medicine, vol. 7, no. 2, 2023.
Jameson, Tanya, and Michael Ford. "MLOps in Healthcare Deployment." Health Systems Engineering Journal, vol. 19, no. 2, 2022.
Gruber, Hans. "Versioning of ML Models in Real-time Environments." IEEE Software Engineering in Health, vol. 11, no. 3, 2021.
Morales, Nina, et al. "CI/CD Pipeline Adaptations for Clinical ML Systems." Journal of Medical Systems, vol. 45, no. 6, 2023.
Watanabe, Yuki. "Containerization of ML in Regulatory Healthcare." AI in Healthcare Research, vol. 9, no. 2, 2022.
O'Connor, David. "Audit Logging in MLOps Pipelines." Healthcare Computing Review, vol. 16, no. 5, 2021.
Kavuri, S. (2023). Machine learning approaches for security vulnerability detection in software testing. Computer Fraud & Security, 2023(10), 21–31.
Lee, Sam, and Rachel Cho. "Deploying Deep Learning Models in Oncology." Journal of Cancer Informatics, vol. 29, no. 1, 2023.
Peterson, Laura. "Data Governance in ML Workflows." Ethics and AI in Medicine, vol. 12, no. 4, 2022.
Ali, Faizan. "HIPAA-Compliant CI for Predictive Analytics." Healthcare IT Journal, vol. 20, no. 2, 2023.
Tran, Kim, and Joel Smith. "Secure Deployment Pipelines for EHR Models." Journal of Health Informatics, vol. 17, no. 3, 2022.
Gupta, Nisha, et al. "Evaluation Metrics for ML Deployment Pipelines." Applied Clinical AI, vol. 5, no. 4, 2024.
Doyle, Patrick. "CI/CD Integration in Federated Healthcare Models." Federated AI in Medicine, vol. 3, no. 2, 2024.
