Research Publication in JMIR Medical Informatics

🔍 Research Overview

• The study, titled “Privacy‑Preserving Prediction of Postoperative Mortality in Multi‑Institutional Data: Development and Usability Study,” was co‑authored by:
  

• Dr. I‑gam Lee and Director Seol‑gi Kim from CryptoLab’s Healthcare AI team,
  

• Prof. Jun‑gyo Seo (Urology) and Prof. Sang‑uk Lee (Anesthesiology) from Seoul Asan Medical Center
  

 Study Goal & Innovation

• Addressed a key challenge: enabling privacy‑preserving collaborative AI across multiple hospitals by securely integrating clinical data encrypted via homomorphic encryption (HE).
  

• Demonstrated that HE-enabled multi‑institutional predictive models outperform single‑institution models, particularly in predicting 30‑day postoperative mortality .
  

 Key Findings & Implications

• Integrated encrypted datasets from multiple hospitals improved model accuracy, addressing data scarcity in smaller clinics and rare disease research.
  

• The methodology enables secure, multi‑hospital data collaboration, minimizing risk of sensitive data leakage while maximizing clinical utility.
  

Researcher Remarks

• Dr. I‑gam Lee (first author):
   “We proposed a way to simultaneously solve the conflicting tasks of acquiring high‑quality medical data and protecting patient privacy. We hope this study becomes the foundation for secure data sharing and utilization among medical institutions.”
  

• Prof. Sang‑uk Lee (corresponding author):
  “This study demonstrates that homomorphic encryption can be applied to actual medical data, opening a new horizon for medical big data utilization. We plan to expand the scope and depth of research through collaboration with more healthcare institutions.”
  
  

Collaboration & Recognition

• CryptoLab and Seoul Asan Medical Center have been collaborating since 2022 on HE applications in healthcare.
  

• Their work earned Best Paper Award at the 2023 Spring Conference of the Korean Society of Medical Informatics and recognition as a top supplier in an AI voucher program by Korea’s National IT Promotion Agency (NIPA).
  

Significance & Future Outlook

• This study showcases a breakthrough in encrypted AI for healthcare, allowing robust, collaborative predictive modeling without compromising privacy.
  

• Widespread adoption could both protect patient data and elevate the quality of medical care, especially in precision medicine and institutions with limited data cryptolab.co.kr.

In summary, CryptoLab and Seoul Asan’s joint research demonstrates that homomorphic encryption can effectively balance data privacy with AI model performance across hospitals—offering a vital path forward for secure, collaborative healthcare analytics.