Blog

FDA Annotation Case Study: How Centaur Labs helped VUNO accelerate FDA clearance for its brain MRI segmentation model

Ali Devaney, Marketing
May 16, 2024

Accurate, on-time medical data annotation by Centaur Labs was the critical piece VUNO needed to land FDA clearance.

Slowly losing a loved one to dementia is a devastating event, but early detection can significantly improve the quality of life for patients and caregivers alike. Dementia and neurodegenerative diseases are prevalent, often untreated, and not only hard to detect early but also difficult to diagnose properly. 

South Korean medical AI company VUNO developed an AI model to aid in the early detection of these conditions and partnered with Centaur Labs to accelerate their FDA clearance. Read on to learn about our FDA annotation services and how our expert radiologists helped accelerate VUNO’s FDA clearance to bring its life-changing technology to the United States.

About VUNO

VUNO, a leading medical AI software company founded in 2014 and based in Seoul, has developed a dozen AI solutions cleared by various countries including EU and Japanese regulators. In 2018 VUNO received clearance from Korean regulators for its product VUNO Med–BoneAge - becoming the first medical AI device ever cleared in the country. Their AI solutions support clinicians in a variety of clinical areas, ranging from analysis of medical images to monitoring patient vital signs. 

VUNO Med–DeepBrain is an AI model designed to aid in the early detection of dementia and neurodegenerative diseases. DeepBrain segments brain regions using brain MRI data and quantifies the volumes of each region to provide atrophy volumetrics against a normative database and generate an analysis report. It received CEMark in 2020.

VUNO’s Challenge: FDA Clearance with a Hard Deadline

Brain images are notoriously complex and time-consuming to read. In an environment where there’s a shortage of radiologists, this created a pressure point for the developers at VUNO. 

When layering in the complexity of a stringent FDA clearance process, the geographical distance, and limited connections with US board-certified radiologists, the task became much bigger for the VUNO team. Through our strategic partnership with Segmed, Centaur Labs was recommended to extend FDA annotation services and accelerate VUNO’s FDA clearance in time for them to launch DeepBrain at the high-profile RSNA conference. 

“We worked with Segmed to collect diverse brain imaging studies and asked them if they were open to helping us find some annotators for our FDA clearance. They said Centaur Labs is the one that you're looking for.”

For prior annotation projects, VUNO had worked with researchers at medical centers in Korea to get datasets labeled. But when preparing for the FDA submission, they needed US board-certified radiologists to annotate a new validation dataset their model had never seen or been trained on to validate the model’s accuracy.

Our Solution: Centaur Labs FDA Annotation Services

Centaur Labs leveraged its network of experts to quickly recruit 3 US board-certified and fellowship-trained neuroradiologists from leading US medical institutions. Centaur Labs and VUNO together trained and managed the annotators on the task based on the protocol outlined by VUNO. The annotation methodology was a double-blind with adjudication approach to quality control the annotations. 

“Brain images were subdivided into about 104 different regions. And, although our AI development team has had a lot of experience working with clinicians, their expertise is in AI and not in brain anatomy. So there's no way we could do the quality control ourselves.”

VUNO pre-processed the data, applying draft masks on the 104 brain regions of interest. The quality of the draft annotations was intentionally designed to be variable to keep the neurorads engaged. Based on those initial draft masks, annotators refined the masks according to their individual experiences. The adjudicator then looked at both labels and either chose one or modified a combination of the two masks to produce the final label.

The Result: FDA Clearance with Time to Spare

All of the required data annotations were complete within 1 month of kickoff and VUNO DeepBrain achieved FDA clearance ahead of schedule. 

“Both of the initial labelers as well as the adjudicator were all very prompt in their labeling. We expected the project to go longer, but we actually finished very much on time, so we were very surprised at the time.”

Through our FDA annotation services, we helped VUNO hit its target go-to-market launch and set the pace for DeepBrain’s US expansion. Our success solidified an ongoing collaboration with VUNO as they work to bring more of their AI models to the US market. 

FDA clearance is just the beginning for global organizations like VUNO. When new medical AI models are developed and deployed, the ripple effect in healthcare can reshape the way diseases are diagnosed and treated. The VUNO team is closer to its goal of improving the declining quality of life experienced by many dementia patients. We remain focused on our mission to support the advancement of new technologies like DeepBrain with accurate, scalable medical data annotations by our crowd of 50k+ medical experts. 

Learn more about Segmed’s real-world imaging solutions for innovation in healthcare.

Get in touch with us to learn more about our FDA annotation services and how we’re enabling the next generation of medical AI developers.  

Related posts