Landmark study pits AI against humans for dental diagnostics


WEST HOLLYWOOD, CA., September 2, 2020 /PRNewswire/ — pearl today announced the results of a landmark study comparing the diagnostic performance of three experienced human dentists to the performance of an artificial intelligence (AI) diagnostic system. The study evaluating the concordance between human and AI scans of a set of 8,767 bitewing and periapical X-rays found that the AI ​​system was more consistent and accurate in predicting the presence of dental caries.

In addition to validating the performance of AI diagnostic systems for dental radiology, the study revealed a troubling lack of diagnostic consensus among human dentists. While they showed moderate alignment – 79% unanimous agreement – regarding the absence of caries, the three human dentists unanimously agreed on the presence of caries in only 370 x-rays – only 4, 2% of total. In nearly one in five cases, even when two dentists identified a cavity on an X-ray, the third dentist did not. The diagnostic variation evident across such a massive radiographic data set is concerning.

“Our intent in producing this study was simply to demonstrate the effectiveness of computer vision machine learning diagnostics in dental radiology, but the secondary findings dig into a major gap in dental healthcare,” said the CEO of Pearl. Ophir Tanz. “While diagnostic inconsistency may be a natural byproduct of human professionals making case-by-case judgments, large-scale inconsistency in standard of care results in suboptimal treatment of patients with wider implications. broad for the health of the population as a whole. Fortunately, the study results also point to a solution: the diagnostic performance of the machine shows that AI is able to infuse coherence into the foundation of dental care. . »

The superior accuracy of the AI ​​system’s diagnostic conclusions remained consistent both when a single dentist’s diagnosis was taken as “ground truth” and when the ground truth came from two dentists. AI also tended to agree more with human dentists than human dentists with each other, a product of AI’s greater sensitivity to potential x-ray degradation.

As a machine-learning system, AI’s intelligence comes from studying millions of x-rays evaluated by hundreds of human dentists. Since the AI ​​brings the information of all these dentists to its scan, when a cavity is present, the AI ​​is more likely to identify it.

Acknowledging the limitations of the study, the authors say they would like future studies to assess the accuracy of more specific diagnostic findings, such as level of decomposition, as well as increasing the number of human participants.

dr. Sanjay Mallyachairman and associate professor of the Oral and Maxillofacial Radiology Section of the Division of Diagnostic and Surgical Sciences at UCLA School of Dentistry, agrees.

“More studies like this are in order,” Dr. Mallya said. “It would be useful, for example, to dig deeper to see if the gap evident in this study persists for caries at different degrees of severity. Is the gap between caries that only affect the enamel, dentin, etc. The more data we have to give credence to the merits of AI diagnostics, because it is this type of data that will help physicians embrace AI as a tool to bring greater excellence and confidence in the care of their patients.

Until that happens, the study published today stands as the first truly compelling, data-driven glimpse into a brave new dental future.

To read the full study – Can a computer identify carious lesions in dental x-rays as accurately as humans? An exploratory study comparing diagnostic assessments performed by humans and a specialized computer vision system – please visit

About the pearl
Pearl is shaping the future of dentistry by delivering AI and computer vision solutions that improve efficiency, accuracy, transparency and patient care. Founded in 2019 by Ophir Tanz and dr. Kyle Stanley, DDS, Pearl is backed by Craft Ventures and several strategic dental industry partners. For more information or to request a demo, please visit

Media Contact
Nate Hermes Where Monique Miller
[email protected]


Related links


About Author

Comments are closed.