AI tool reveals six distinct craniofacial patterns underlying one of the most challenging conditions in orthodontics: mandibular prognathism

Researchers from the ERA Chair iSTARS Project, Prof. João Guimarães and Dr. Inês Carvalho, together with Dr. Maria Cristina de Faria Teixeira from Lisbon Medical School (FMUL), led the development of an artificial intelligence model capable of stratifying patients diagnosed with skeletal malocclusion Class III—one of the most complex orthodontic conditions to treat—into six distinct subtypes.

The study, conducted in collaboration with Prof. Alejandro Linares from the School of Dentistry at the Complutense University of Madrid, was published in the prestigious scientific journal Nature Communications Medicine and represents a significant advance in diagnosis and personalized treatment planning in orthodontics.

Skeletal malocclusion Class III occurs when the mandible projects forward relative to the maxilla, giving the face a concave profile, and is commonly referred to as prognathism. The condition affects both adults and growing patients, with a prevalence of approximately 3.4% in Europe, and can lead to functional and psychological challenges.

“Skeletal malocclusion Class III is one of the most difficult malocclusions to treat and represents one of the greatest clinical challenges in orthodontics and dentofacial orthopedics. Organizing these patients into phenotypic subgroups allows clinicians to better identify dysmorphic structures and clarify the severity of these discrepancies,” explains Dr. Maria Cristina de Faria Teixeira from Lisbon Medical School.

Until now, orthodontists lacked a standardized way to classify the many variations of skeletal malocclusion Class III, which made treatment planning difficult and sometimes inconsistent.

“Our goal was to develop a data-driven artificial intelligence model capable of identifying morphologically consistent subphenotypes, in order to support a more objective diagnosis and, consequently, better personalize treatment,” adds Dr. Inês Carvalho, data scientist in the iSTARS project at Lisbon Medical School.

The research team analyzed cephalometric radiographs from 655 adult patients and mapped twelve facial anatomical landmarks. Using advanced geometric morphometric techniques and machine learning algorithms, they grouped patients into six distinct subtypes based on subtle differences in craniofacial structures.

“Our study results contribute to a more comprehensive analysis and understanding of this condition, thanks to the pioneering inclusion of populations from different ethnic backgrounds within a single study,” highlights Dr. Maria Cristina de Faria Teixeira.

“One of our surprising findings was that the groups defined solely from anatomical data also reflected distinct treatment approaches. In particular, we identified phenotypic subgroups associated with different levels of severity and with the need, or not, for surgical intervention,” adds Dr. Inês Carvalho.

To support the adoption of this new technology and foster further research in the field, the team has also made available an online application that allows clinicians to input a patient’s anatomical landmarks and obtain the corresponding subphenotype.

By providing a more precise understanding of each patient’s unique craniofacial structure, the model may help orthodontists select the most effective treatment plan from the outset, reducing unnecessary procedures and improving clinical outcomes.

The researchers note that next steps include pilot studies in clinical settings and the integration of 3D radiological data, which may further enhance the model’s accuracy.

About the Study
The article, entitled “Geometric morphometrics based diagnostic model for Skeletal Class III patients”, is available in Nature Communications Medicine, and the online tool can be accessed at: https://tools.istars.pt/sciii/

Full article available at: https://www.nature.com/articles/s43856-026-01557-y