Connect with us

Hi, what are you looking for?

Artificial Intelligence

Centaur Labs Gets $15 Million To Improve Data For Healthcare AI

The goal is to improve the quality of the data used to train AI programs and build the tools and processes required to put data at the center of developers’ work.

Centaur Labs Gets $15 Million To Improve Data For Healthcare AI

Approximately 30% of the world’s data volume is created by the healthcare industry, according to RBC and IDC. The compound annual growth rate of healthcare data between 2018 and 2025 is predicted to be 36%, a much faster rate of data growth than that of other industries, including financial services at 26%.

Last year, healthcare’s share of all the data created worldwide amounted to 21 zettabytes or 21 trillion gigabytes. Also last year, Covid-19 made all of us aware of the important role of high-quality data in the successful enlistment of AI in humanity’s battle with diseases and in keeping people healthy.

Centaur Labs, a startup focused on improving the quality of healthcare data, today announced $15 million in funding to advance their mission to label the world’s medical data. The Series A round was led by Matrix Partners with participation from other funds including Accel, Global Founders Capital, Susa Ventures, Y Combinator, and individual investors.

Today’s “artificial intelligence” is the new generation of machine learning.  This new, “deep learning,” approach is based on highly sophisticated statistical analysis of very large volumes of data, “training” machines to distinguish between good and bad, positive and negative, illness or wellness.

The first step in the training process is to present labeled data to the computer program as examples of what’s right and what’s wrong so that these type of computer programs (or algorithms) could make accurate classifications of non-labeled data. Bad data, however, can lead to bad diagnosis, decisions and outcomes. The efficacy of these algorithms, their potential for improving health and healthcare, largely depends on the accuracy of the underlying data labels.

Centaur has assembled a network of tens of thousands of medical students and professionals from over 140 countries. This network primarily labels data on Centaur’s gamified iOS app, DiagnosUs, where labelers improve their skills and compete with one another. The app is designed to judge labelers on their performance and reward the most accurate labelers with cash prizes. Importantly, Centaur collects multiple opinions on every case—with more opinions collected on the most difficult cases—and intelligently combines those opinions into labels that are more accurate than those from an individual expert. More than 1 million opinions are contributed through the platform each week.

The work that Centaur Labs is doing and its focus on the quality of healthcare data is in line with AI pioneer’s Andrew Ng recent campaign to shift AI development from being model-centric to being data-centric. The goal is to improve the quality of the data used to train AI programs and build the tools and processes required to put data at the center of developers’ work.

“Now that the models have advanced to a certain point, we got to make the data work as well,” Ng told me recently. And writing in the Harvard Business Review, Ng advocated “focusing on data that covers important cases and is consistently labeled, so that the AI can learn from this data what it is supposed to do… the key to creating these valuable AI systems is… teams that can program with data rather than program with code.”

Centaur Labs is focused on making data, specifically healthcare data, work very well. It was founded by Erik Duhaime, CEO, while he was a PhD student at the MIT Center for Collective Intelligence. Other founders include his long-time friend from Brown University, CTO Zach Rausnitz, and VP of Engineering Tom Gellatly, who managed the data labeling team at the self-driving car company Cruise Automation and previously was the Head of Mobile Development at the ridesharing startup Sidecar.

“AI learns like humans—by example—and to train an algorithm it takes thousands or even millions of examples. It is difficult to curate large medical datasets, and nearly impossible to source accurate labels from those with medical knowledge and specialized training. Our platform is built to support a wide range of specialized medical tasks, and to quickly scale to millions of labels,” co-founder and CEO Erik Duhaime said in a statement.

Seen on Forbes (Innovation): Article Link

Click to comment

Leave a Reply

Your email address will not be published. Required fields are marked *

Latest News


Dr Ishara Dharmasena, of the School of Mechanical, Electrical and Manufacturing Engineering (MEME), has been appointed a research fellow by the Royal Academy of Engineering and awarded funding...


Evidence of the impact on medtech emerged over the summer, when ResMed cited the shortage of semiconductors as a reason it will be unable...

Consumer Empowered Health

Meru Health is a startup that aims to solve a number of the challenges associated with getting access to quality mental health care, with...


The NHS has launched the world’s largest trial for a blood test that can detect 50 types of cancer before symptoms appear.

You May Also Like

Artificial Intelligence

Digital Solutions In Healthcare Plug “digital healthcare solutions” into Google, and the results are clear: digital tools have taken the healthcare industry by storm....


In Delray Beach, Florida, an emerging biopharma company, Cyrano Therapeutics, is planning a clinical trial using approved theophylline via a new intranasal pathway of...

Artificial Intelligence

Joe Mullings discusses the Microbot – Liberty robotic platform with Harel Gadot and how a new category of surgical robotics creates a new entry...

Digital Health

Software as a Medical Device (SaMD) is driving innovation in the HealthTech space and is more popular than ever — completely upending the way...