Dental care is experiencing a quiet but dramatic technological development. Many of the advances in artificial intelligence and big data that are changing the way we work, shop and find entertainment may soon make it easier for dentists to customize care, monitor patients and develop new treatment options.
UIC College of Dentistry researchers are among the first in the country to explore the value of 3D image analysis, data sharing on the blockchain, home monitoring and other clinical technologies. Their perspective is not simply to chase the latest hyped technology, but to determine how these methods can improve patient outcomes and access to dental and orthodontic care.
UIC dentists and orthodontists often treat patients with rare or complex conditions, including cleft palate, ectodermal dysplasia, craniosynostosis, and other complex dentofacial deformities, providing dental care to traditionally underserved populations. These activities provide an opportunity to develop AI concepts and rigorously test whether they provide meaningful benefits compared to current methods and avoid perpetuating bias.
“It’s a huge advantage to be in an urban area and serve a population that is traditionally disenfranchised and underrepresented,” said Dr. Veerasathpurush Allareddy, Brodie Craniofacial Chair and Professor of Orthodontics at UIC. “This means that when we train these AI models, we can better adjust for some of the unique factors or challenges that these groups of people face and ensure more algorithmic fairness.”
AI for diagnosis, treatment and remote monitoring
Clinicians’ use of X-rays and other scans is essential for determining treatment plans and monitoring patient outcomes, but repeatedly collecting and analyzing these images is time-consuming for both physician and patient.
Artificial intelligence offers new approaches to exploiting this important visual data. UIC dentists are collaborating with colleagues in the College of Engineering to develop new algorithms for analyzing these images; for example, determining the amount of growth left in a patient to guide orthodontic treatment. Clinicians currently use a broad four-category scale to measure this development and choose between surgery and other interventions, but a new model provides a finer, continuous measurement.
“We used image analysis, image processing and deep learning methods to assess the maturity of a patient from spinal cord X-rays,” said Ahmet Enis Cetin, professor of electrical and computer engineering. “This will be the first step towards a personalized approach to surgery, where AI is used as a new tool to help dentists make decisions.”
A team of dentists and engineers led by Dr. Mohammed Elnagar, assistant professor of orthodontics, also created an AI algorithm that helps choose the most effective treatment plan. For example, the model – trained on 18 years of patient records collected at the dental college – can judge whether a patient’s treatment goals can best be achieved by using only braces, or whether additional surgical procedures are required. Members of the team received Thomas M. Graber Awards of Special Merit from the American Association of Orthodontists for the work.
Once treatment has started, artificial intelligence can also help clinicians monitor the patient’s progress, even from afar. For example, patients can use attachments to their smartphone to take their own oral scans at home. An AI algorithm then creates a 3D model that shows how the treatment is progressing and identifies potential problems that may require an office visit.
In recent papers, UIC researchers found that the quality of these home scans matched what was achieved with regular clinical scans. The same performance is encouraging for using the technology for early detection of complications. An ongoing clinical trial is testing whether treatment decisions guided by remote monitoring technology are as effective as in-person care.
“You can optimize the patient’s office visit based on the individual response,” Elnagar said. “If they respond, they can continue without a visit. If it moves off track or there’s a surprise, we can have them come earlier.”
The ability to collect high-quality scans at home will also make a meaningful difference for patients who live far from clinics or with complicated conditions that currently require frequent visits.
“It’s really exciting for patients who are limited in their choice of providers,” said Dr. Min Kyeong Lee, a clinical assistant professor of orthodontics who also studies the ethical considerations of AI applications. “It can save a lot of time traveling, and at the start of some treatments it can sometimes require a visit every month, so it’s a big burden on the family that we can reduce.”
Genomics and blockchain
While imaging combined with artificial intelligence can provide powerful data about a patient’s current condition, dentists also want to be able to make accurate predictions about their future. To do so will require additional data, including a rich category that has already made a massive impact on medicine: genomics.
In orthodontics, procedures can last for years, and clinicians must anticipate how a patient’s teeth and associated structures will change in order to find the most appropriate treatment option. Genomic information could help remove this uncertainty and inform decisions, Allareddy said, by identifying associations between certain genes and factors such as root resorption and tooth movement.
“We will be able to provide truly personalized orthodontic care based on the genomic profile of each patient,” Allareddy said. “We can change the treatment or maybe even stop doing the treatment when the data suggests that we would probably do more harm than good.”
But dentistry does not have the same access to universal platforms or the culture of data sharing between institutions that exists in medicine. While the UIC College of Dentistry has the advantage of being one of the largest programs in the United States, the types of studies that will unlock the predictive capabilities of genetics will require data from much larger patient pools combined with other forms of information, including images and clinical outcomes.
A potential facilitator of these necessary data exchanges could be blockchain. Although often discussed in the context of cryptocurrency, blockchain also offers promise as a secure record of information distributed across computers worldwide rather than a single centralized database.
In a recent paper, Allareddy, Elnagar, Lee and Dr. Maysaa Oubaidin, associate professor of orthodontics, to use blockchain technologies to help dental researchers around the world share and learn from clinical data. The system could enable federated machine learning – a form of artificial intelligence where models are trained on distributed data without moving it from its safe home – and large-scale analytics, unlocking the potential for genomic data or testing interventions across a wider range of individuals, including . those from underserved patient populations.
“If we can really have this huge collaboration between universities, hospitals and clinics, and we can focus on organizing the data so that everyone can work on a big project together instead of competing, then it might be possible to have some bias or exclusion of minorities,” said Dr. Flavio José Castelli Sanchez, an assistant professor of orthodontics. “We’re trying to eliminate that, and there’s a great chance to do it with artificial intelligence, as long as we do it properly.”