Deepnoid Partners with Furiosa AI to Enhance Generative Medical AI
Translated from Korean, summarized and contextualized by DistantNews.
At a glance
- Medical AI company Deepnoid has launched its generative AI-based digital medical device, M4CXR, and is collaborating with Furiosa AI.
- The company also introduced its medical AI foundation model, MedZero, which analyzes various medical images and integrates with EMR and OCS systems.
- M4CXR has received regulatory approval as an innovative medical device and a digital medical device, aiming to alleviate the workload of radiologists by automating draft reports.
South Korean medical AI specialist Deepnoid is expanding its reach in the healthcare sector with the introduction of its generative AI-powered digital medical device, M4CXR, and its specialized foundation model, MedZero. The company announced these advancements at a press conference, highlighting collaborations with Furiosa AI to enhance business viability. Deepnoid, established in 2008 and listed on KOSDAQ in 2021, currently offers solutions for medical image interpretation and diagnostic assistance, including DEEP:NEURO and DEEP:CHEST.
In Korea, there are about 300,000 medical readings per year, and the number of doctors to read them is absolutely insufficient. We are providing various AI to automate tasks that can be applied to this part.
M4CXR has achieved a significant milestone by becoming the first generative AI medical device to be designated as an innovative medical device, followed by obtaining digital medical device approval. The device aims to address the overwhelming workload faced by radiologists. According to Dr. Kim Sung-hyun, CEO of Human Medical Imaging Center, radiologists often face over 100 CT scans daily, with a high percentage being normal. M4CXR assists by generating draft reports, allowing doctors to focus on abnormal findings and improving efficiency. Clinical trials showed M4CXR's hallucination rate at approximately 0.2%, with reports comparable to those of second or third-year radiology residents.
The biggest problem faced by radiologists in Korea is the excessive workload. While the medical community considers about 30 CT scans a day to be the maximum reading volume, the actual processing volume exceeds 100.
Deepnoid is also developing MedZero, a multimodal AI foundation model for medical use, set for release by the end of the year. MedZero is designed to analyze diverse medical images like X-rays, CT, and MRIs, and will be integrated into AI agents that connect with EMR and OCS systems. CEO Choi Woo-sik stated that MedZero's performance is superior to Google's Med-PaLM. The company plans to expand its digital medical device approvals based on M4CXR's success. Furiosa AI is providing technical support for the infrastructure, including GPUs and NPUs, to ensure the stable operation of Deepnoid's AI services, particularly in the report generation process for M4CXR.
Out of 100 CT scans, 95% are normal findings. While we should focus on analyzing the 5% of abnormal findings for the patient's sake, the excessive workload affects the analysis of abnormal findings.
Originally published by Dong-A Ilbo in Korean. Translated, summarized, and contextualized by our editorial team with added local perspective. Read our editorial standards.