SuperbAI supplies mobile phone-based dental cavity diagnosis AI model to Ajou University… ‘No plans for a general-purpose model’
SuperbAI, a company specializing in vision AI (CEO Hyunsoo
Kim), announced on the 13th that it developed a mobile phone-based
dental cavity diagnosis AI model in just three weeks and supplied it to Professor
Heekyung Kim’s research team at Ajou University Dental Hospital.
The newly developed dental cavity diagnosis AI model aims
to assist professional diagnoses by detecting cavities early using images taken
with a smartphone. It is expected to be particularly useful for individuals who
have difficulty visiting the dentist due to busy schedules or those living in
areas with limited medical infrastructure.
This project was initiated when Professor Heekyung Kim of
Ajou University Dental Hospital approached SuperbAI for collaboration on
developing an AI model for cavity diagnosis. The partnership began to address
challenges in the AI model development process, including data scarcity,
security concerns, and data standardization.
SuperbAI notably completed the initial AI model within just
three weeks of starting the project. The diagnostic criteria of four
participating dental specialists were incorporated into the model, and its
accuracy was continuously improved through ongoing refinement based on their
feedback.
The limited cavity data was supplemented using generative
AI-based data augmentation techniques. Synthetic data technology is one of
SuperbAI’s core proprietary strengths.
The company added that standardization efforts were also
carried out to minimize the variability in image quality captured by smartphone
cameras. They emphasized that medical data was handled securely through an
advanced internal security system within the platform, while data collection
guidelines were established and efforts to continuously gather insufficient
data were maintained.
As a result, the model achieved high levels of
precision—the rate at which actual cavities were correctly detected—as well as
recall, the proportion of total cavities identified, and the F1-score, which is
the harmonic mean of precision and recall. The company stated that they have
developed a platform capable of delivering real-time diagnostic results,
bringing the technology to a level suitable for practical use in real-world
settings.
Professor Hee Kyung Kim of Ajou University Dental Hospital stated,
“Medical diagnosis involves a degree of subjectivity from the physician, which
makes standardization challenging,” and added, “Thanks to the project manager’s
efforts to learn dental terminology and actively incorporate our feedback, we
were able to achieve a collaborative process without compromising on
professional expertise.”
Hyun Soo Kim, CEO of SuperbAI, stated, “We’re able to build
the first prototype of the desired AI model within just three weeks and
continuously refine it, allowing for rapid performance verification from the
proof-of-concept (PoC) stage before moving on to large-scale projects.” He
added, “We will continue to contribute to the advancement of digital healthcare
technology through our innovative capabilities and platform.”
Meanwhile, the company stated that there are currently no
plans to develop a general-purpose model based on this one that would allow
anyone to perform self-diagnosis.
A company representative said, “There is growing demand in
the field of medical imaging AI, and we’re receiving many inquiries,” adding,
“However, since this is a domain where specialized knowledge is absolutely
essential, we have no plans to independently develop a general-purpose model
for the time being.”
They added, “That said, we’re gaining a lot of valuable
experience through our collaborations in the medical field, and if the
opportunity arises, we plan to continue working with dental professionals.”
댓글
댓글 쓰기