JLK Secures Patent for Medical Imaging and Clinical Text Integrated LLM Technology

JLK Secures Patent for Medical Imaging and Clinical Text Multimodal LLM Technology

JLK secured a patent on the 19th for an AI technology that integrates medical imaging and clinical text data into a multimodal large language model, medigatenews.com reported. The system structures lesion locations, volumes, and vascular occlusion data to reduce input token counts for large language models while preserving clinical meaning.

Multimodal AI platforms increasingly combine text records from electronic medical record and electronic health record systems with image data from picture archiving and communication systems. Inputting large medical files like 3D CT and MRI scans directly into a language model causes exponential token growth and high operating costs. JLK designed its patented system to structure key clinical features and apply a grounding structure based on quantitative analysis values and medical guidelines to limit answer ranges.

Integration with JOOMED and Hardware Verification

The patented technology serves as the core engine for JOOMED, an AI platform developed by JLK that connects with hospital systems to analyze medical images alongside patient test results and examination records. In July, the company tested the platform operating environment using Intel Arc Pro B70 and NVIDIA A100 hardware, confirming consistent analysis results across both setups, medigatenews.com reported.

“To utilize artificial intelligence in actual medical fields, specialized technology that understands medical images and clinical information together and connects them with hospital systems is important,” said Lee Myung-jae, an official at JLK, according to medigatenews.com. The company plans to expand its business into next-generation medical AI platforms by combining traditional stroke imaging AI with this multimodal LLM technology.

Frequently Asked Questions

What specific data does the patented JLK system structure to reduce token counts?

The system structures lesion locations, lesion volumes, and vascular occlusion information from clinical data to minimize input token volume while retaining diagnostic relevance.

How does the platform ensure reliability in its AI-generated answers?

The technology applies a grounding structure that restricts the LLM’s response range using quantitative analysis values and established medical guidelines.

Which hardware configurations were tested with the JOOMED platform?

JLK verified the platform operating environment using Intel Arc Pro B70 and NVIDIA A100 hardware configurations, according to company reports.