Baltimore-based startup Dentuit is working to bring AI to the dentist’s office.
Dentuit caught our attention last year, earning an honorable mention spot on Technical.ly’s RealLIST Startups 2021. A lot has happened, so we got in touch for a check in.
And since founder Dan Lee has plenty to share about building tech skills and startups in Baltimore, so we decided to open up the conversation for an AMA on the Technical.ly Slack. (It’s where the whole Technical.ly team is hanging these days. Get in touch if you want to join us for an AMA at firstname.lastname@example.org.)
Here’s a look at what we learned:
A path through Baltimore tech
Lee first got a taste of tech through a web development bootcamp at Betamore, the Port Covington-based hub of startup incubation and dev education. From there, internships introduced him to the Baltimore startup scene, including at the ETC (Emerging Technology Centers) incubator in Highlandtown. From there, it was off to UMBC for a computer science degree. He chose tech for a career because of the opportunities for creation — both with code and companies.
He decided on computer science as his major “when I started to realize just how much you can create with code, from websites to entire full-stack applications. And then to bring it back to ETC days, the whole startup vibe there with passionate entrepreneurs around you really hooked me in to considering applying these skills into an actual career,” he said. “Timing is also everything … and I think it was around the time where access to knowledge via the internet was only getting better at a rapid pace.”
A founder’s mindset
The early introduction to the startup scene also helped Lee to get an idea of how he wanted to direct his studies. He went in to gain tech skills, but wanted to put them to use for entrepreneurship.
“The mindset,” he said, “is really critical in shaping the value that you get from an education.”
Opportunity arose during his last semester at UMBC. A conversation with his brother, who is a dentist, identified some nagging pain points.
Without mentioning a specific technology, “I literally asked him, ‘If you had access to a software technology that doesn’t yet exist but can really help with some top pain points you deal with day to day, what would it be?’” Lee said. “Then when he said he wished there was some technology that would help him better read his patients’ X-rays, I immediately thought it could be solvable with the advancements being made in medical AI.”
From there, Lee got to work. He took a graduate-level computer vision class, and developed a proof-of-concept to do automated detection of lesions, or, anything abnormal in teeth and gums, in X-rays. With that in hand, he won first prize at the Cangialosi Business Innovation Competition, and used the funding to launch Dentuit. Now, the company is based at Betamore.
Lee has continued to talk to dentists along the way.
“A top concern among dentists is (based on interviews) the likelihood that they missed 1-2 carious lesions when reading an X-ray perhaps too quickly or the image quality is just not ideal,” he said, “then when the dentist examines the patient manually, you’re going to only look at areas of the mouth that you see from the X-ray (and miss the ones you overlooked).”
The company’s product is a tool to support dentists as they make decisions about treatment while reading an X-ray. There’s software on the market now, but it requires a dentist identify an area of concern.
“What makes dental AI in general unique is everything is automated such that the image is analyzed by the model and learns to identify carious features on its own,” Lee said. “No longer are the days of having machine learning practitioners handcraft edge detection algorithms.”
The team has also been careful to design for a group of users that will be new to artificial intelligence.
“From the outset we wanted to make sure any dentist using any imaging software setup can have access to AI support,” he said.
To get to market, it takes proving out the technology first. To do so, Dentuit secured Phase I funding from the National Science Foundation through the Small Business Technology Transfer program.
Such programs can help technologies develop, and it’s a particularly compelling option in the health space where so many Baltimore startups operate. As he seeks capital to continue building, it’s also a reminder that startups are not all built the same. A dental AI startup isn’t like a typical consumer-facing app or SaaS company. Getting into dental offices will take a different kind of path than an app store, and, as with many new technologies, the regulatory picture is still being developed. It means different expectations and time that it will take to get to market, but Lee remains confident that there can still be a good outcome.
Going forward, the company is focused on validating its technology with a data-centric focus and meeting with vendors and investors interested in the space.
What’s your stack?
We asked Lee about his technology stack. For AI development, he responded this way:
“We used Labelbox to help us manage and crowdsource annotations of our dental X-rays among the various clinical volunteers that we managed to recruit,” he said. “Then we use PyTorch as our [machine learning] framework, and NVIDIA’s MONAI framework for specifically training on labeled X-ray images.”
For Dentuit, the tools for day-to-day productivity are just as important.
Lee has long appreciated Trello, and he links the company’s account with Google Drive to keep track of all documents and assets. For weekly planning meetings, they use Google Keep to add and track items over the course of a week.
“This helps us be retrospective in how much we were able to get done, see what worked well, what didn’t, and use that to plan for the upcoming week in optimal manner,” he said.
Lessons from the last 18 months
With the pandemic, the company moved to Zoom for communication, but Lee said the processes in place allowed the team to keep going without changing much. Still, the last year has brought lessons.
“What I’ve learned is that it pays to master the art of communication, be it with good team members or tools that help streamline that,” he said, “and then the importance of balance which I also think is key to not burning out from being online too much because we’re all working from home.”
To maintain that, Lee said that team members have one particular strategy.
“A weird detail is that we don’t communicate via SMS or even share our mobile phone,” he said. “It actually works well since we reserve Zoom chat messages for emergency immediate needs and Trello mentions for noteworthy things to be aware of that do not need immediate attention.”
With more digital tools for workflow comes a chance to optimize each. Some are perhaps best left for personal lives only.
You can join or just follow along with conversations like this one in the #ama channel on our public Slack. Coming on Thursday, Aug. 26, Technical.ly reporters Holly Quinn and Paige Gross will lead a discussion with women-in-tech meetup organizers from across the mid-Atlantic.