Seoul National University Hospital has recently unveiled a clinical decision support system for precision medicine treatment.
WHY IT MATTERS
The SNUH POLARIS (Precision Oncology and Rare-Common Disease Supporter) is said to be the first of its kind in South Korea that supports personalised treatment driven by AI and based on big clinical and genomic data.
Powered by a large language model (LLM), it extracts, integrates, and refines previously scattered data on pathology, diagnostics, genomes, surgery, and treatment information from the HIS.
It also incorporates the hospital’s next-generation sequencing cancer panel to quickly provide real-time comparison and analysis of big genomic data, informing cancer diagnosis at the point of care. The resulting insights are also cross-validated by clinical and genomic experts.
SNUH POLARIS, developed by a team comprising 30 departments and 87 multidisciplinary staff, is initially applied to cancer cases and will later be expanded to cover rare diseases and chronic diseases.
THE LARGER TREND
SNUH recently discarded its legacy IT systems and transitioned to a private cloud and SDN-based environment to simplify and minimise physical network configurations. It also adopted advanced encryption, multi-factor authentication, network and endpoint security, real-time monitoring, and disaster recovery systems, as well as acquired the national information security certification from the Korean government, to fully ensure the protection of patient data, operational continuity, and regulatory compliance. These measures have been recognised in its recent validation for Stage 6 of the globally recognised HIMSS Infrastructure Adoption Model.
With a robust IT infrastructure in place, the hospital is prepared to pursue more big data projects. Recently, SNUH and its two affiliate hospitals in Bundang and Boramae have consolidated their common databases to enable network-wide studies and collaborations.
Meanwhile, SNUH introduced in April what could be the first medical LLM in South Korea. While it has yet to be formally used to assist clinical research, a study showed that the model outperformed takers of the Korean Medical Licensing Examination and could process 50,000-word texts simultaneously.
