JLK, a medical artificial intelligence (AI) specialist firm, announced on Monday its plans to target the integration of medical imaging and internal hospital text data through its multimodal large language model (LLM) platform, JOOMED.
JOOMED is an LLM platform capable of automatically analyzing high-dimensional medical imaging data, such as CT and MRI scans, and can be used for similar case searches and prognostic research.
JLK positions JOOMED strategically as securing a foothold in the medical imaging data layer, an area challenging for general-purpose LLMs to penetrate.
Rather than directly competing with general AI models like ChatGPT, Claude, or Gemini, JOOMED is described as an essential infrastructure platform for processing, interpreting, and structuring medical imaging data – a crucial step for deploying general LLMs in healthcare settings.
JLK’s medical imaging analysis engine selects, analyzes, and structures key information from images, transforming it into a format suitable for LLM input. This process compresses the input cost of brain perfusion CT images (CTP) from about 870,000 tokens to just 4,500 tokens.
By reducing token usage and computational load, JOOMED focuses on compiling reference materials that allow healthcare professionals and researchers to review medical images, clinical information, related literature, and guidelines collectively.
JOOMED’s scalability shines in similar data searches and prognostic research using internal hospital data. Its core differentiating feature is the ability to search for synchronized similar data between medical images and various clinical text information, providing relevant insights. This goes beyond simple keyword searches, calculating multidimensional similarity by comprehensively considering imaging findings, clinical information, treatment progress, and prognostic data.
A JLK spokesperson explained that JOOMED isn’t designed to compete with general LLMs. Instead, it’s an integrated multimodal LLM platform based on specialized medical imaging data, crucial for LLMs to be effectively utilized in healthcare settings. The medical images, reports, treatment outcomes, and prognostic data within hospitals represent a private data domain that external general models struggle to access, requiring diverse and complex algorithms.
In related news, JLK has obtained the SOC 2 Type II certification, a U.S. information security standard, and has also received third-party verification for compliance with Health Insurance Portability and Accountability Act (HIPAA) regulations. This achievement bolsters JLK’s information security credibility as it expands its business targeting U.S. healthcare institutions.
SOC 2 Type II is a prominent certification that evaluates the security operational capabilities of cloud-based services and digital healthcare companies. It goes beyond merely checking for the existence of policies, verifying that a consistent security control system operates effectively over a specific period.
Through a rigorous three-month audit process, JLK demonstrated high-level operational capabilities in key information security areas, including data security, access control, system operational stability, and internal management procedures.