
The researchers of technical schools of Tehran University have developed an artificial intelligence system that is able to diagnose common diseases such as caries, gum disease and dental impaction with high speed and accuracy by analyzing panoramic images of teeth. This tool is supposed to play the role of an intelligent assistant for dentists and revolutionize the process of diagnosis and treatment planning.
According to Ecomotive, citing Mehr, this scientific achievement is the result of Mahdieh Dehghani’s doctoral dissertation under the supervision of Reza Aghaizadeh Zarebi, a professor at the Faculty of Electrical and Computer Engineering, and its results have been published in an international journal. The researchers say that the developed model can automatically determine the exact location of lesions and abnormalities and provide the probability of correct diagnosis.
Explaining the importance of this research, Reza Aghaizade Zarebi says that X-ray panoramic images are one of the most important diagnostic tools in dentistry; But their manual interpretation is time-consuming and its accuracy depends on the experience and working conditions of the doctor. He emphasizes that the increase in data volume and work pressure can increase the possibility of human error, and artificial intelligence can play the role of an effective auxiliary tool in such conditions.

According to him, this system is developed based on YOLOv11 advanced architecture and trained with a public dataset. The new model not only extracts lesions and abnormalities with high accuracy, but also declares the confidence level of the system in its diagnosis; A topic that can be very valuable for clinical decision making.
One of the strengths of this project is its implementation in the form of a web application so that dentists and radiologists can use it without the need for special equipment. By identifying problem areas in the image, this tool helps clinicians prioritize faster and increase reporting accuracy.
Researchers say that compared to previous versions, the proposed model has a more accurate performance in identifying and classifying anomalies and can reduce errors caused by fatigue or high workload as a reliable “second observer”.
RCO NEWS



