Deepnoid is embarking on the commercialization of generative medical artificial intelligence (AI) with the recent approval of its chest X-ray interpretation solution, 'M4CXR.' CEO Choi Woo-sik announced this initiative during the '2026 Deepnoid Media Day' held at the Westin Chosun Hotel in Seoul.
Choi stated, "The approval of 'M4CXR' marks the starting point for generative medical AI to provide tangible benefits in hospital settings, transforming us into a 'medical AI service company' that encompasses solutions, infrastructure, and agents." He emphasized that Deepnoid will no longer define itself solely as a medical device company, but will expand its modalities to include chest CT and MRI.
'M4CXR' is the first generative AI medical device to receive designation as an innovative medical device and has now obtained product approval from the Ministry of Food and Drug Safety.
According to the company, 'M4CXR' has been trained on over 10 million chest X-ray images and reports, enabling it to interpret more than 41 abnormal findings and automatically generate preliminary reports in an average of 2.3 seconds. Deepnoid aims to address structural issues in the medical field, such as the shortage of radiologists, the concentration of interpretation infrastructure in large hospitals, and the workload from repetitive chest imaging.
Additionally, the company is leveraging its accumulated experience in chest and brain imaging products, along with a nationwide hospital network and partnerships with Picture Archiving and Communication Systems (PACS), to establish a foothold in the market.
Starting with 'M4CXR,' Deepnoid plans to expand its modalities. The roadmap includes extending chest imaging from X-ray to CT and MRI, and broadening brain imaging from MRA to CTA and MRI, referred to as a 'dual-axis expansion' strategy. Choi expressed his vision to evolve beyond simple diagnostic assistance to an agent-based AI service that aids in generating preliminary reports and executing workflows in medical settings.
The event also featured a medical session that validated the clinical utility of 'M4CXR.' Kim Sung-hyun, director of the Human Imaging Center, explained that 'M4CXR' automatically generates preliminary reports, allowing specialists to focus on identifying abnormal findings. He described it as a "smart assistant role, not a replacement."
In the clinical efficacy evaluation presented at the briefing, 'M4CXR' demonstrated a suitability rate of 96.6%, closely matching the 97.6% suitability rate of a radiologist with over ten years of experience, confirming its high-quality interpretation capabilities.
On the infrastructure front, the collaboration with FuriosaAI is noteworthy. The two companies are conducting a demonstration combining 'M4CXR' with FuriosaAI's Renegade (RNGD) chip through the National IT Industry Promotion Agency's (NIPA) 'AI Semiconductor Application Demonstration Support Project,' achieving over double the performance per watt compared to NVIDIA's H100.
A Deepnoid representative stated, "We believe that true commercialization is only possible with a stable operational infrastructure and will accelerate the expansion of a domestic technology-based medical AI ecosystem."
* This article has been translated by AI.
Copyright ⓒ Aju Press All rights reserved.