Cyber Dentistry: AI Conquers Nerve Canal Detection in 3D X-rays

Posted by

In the dystopic cityscape of the not-so-distant future, the meticulous application of cutting-edge artificial intelligence technology becomes essential to everyday life. Take, for instance, the humble role of a dentist preparing to insert a tooth implant into a patient’s lower jaw. To ensure the success of the procedure, the precise location of the nerve canal must be known in order to plan the implant’s size and position, as well as the overall execution. Currently, this critical information is extracted from X-ray images, requiring the dentist or radiologist to manually identify the canal location point by point. This process can be tedious, time-consuming, and ultimately, imperfect.

As the need for speed and accuracy in the dental field increases, Planmeca, a dental equipment manufacturer, has teamed up with the Finnish Center for Artificial Intelligence (CAI) and Tampere University Hospital (Tays) to develop a more efficient, technologically advanced solution. Their breakthrough innovation – an AI-based model capable of identifying the lower jaw nerve canal in 3D X-rays more swiftly and accurately than its human counterparts or any other existing automated methods.

This groundbreaking collaboration was born from clinical practitioners seeking ways to streamline their workflow and utilize artificial intelligence to save time in treatment planning. Vesa Varjonen, Vice President of Research and Technology at Planmeca, sees this as the inevitable step forward in the integration of technology into healthcare, firmly stating that their AI application offers a more efficient route.

Employing deep neural networks to analyze a vast collection of clinical data, which includes three-dimensional images captured with cone beam computed tomography (CBCT), the new method transcends its organic counterparts with ease. Tampere University Hospital provided invaluable input to this project, supplying extensive and diverse clinical materials gathered using multiple 3D-imaging devices.

Jaakko Sahlsten, a doctoral researcher at Aalto University, describes the division of this data, explaining how it was categorically allocated for the dual purpose of training the neural networks and for testing and validating the final devised method. The synthesis of artificial intelligence and clinical practice has become an efficient and reliable tool in this modern era.

Nerves controlling jaw movement and facial senses are located in the mandibular canal, making it a vital component to consider during implant placement, wisdom teeth extraction, or jaw surgery. As the location and path of the canal within the jawbone are unique to each individual, the AI model faced challenges caused by the minuscule size of the mandibular canal in comparison to overall image data. Sahlsten refers to the training material dataset as ‘highly unbalanced’ due to this data asymmetry.

Working closely with Tays radiologists proved to be instrumental in utilizing data effectively during the AI training process. According to Varjonen, the neural network learns to optimize its internal parameters when vast amounts of data are inputted and the mandibular canal’s location is marked. The end result – a neural network capable of rapidly identifying an individual’s mandibular canal from a single 3D data input.

Upon testing the neural network model with isolated patient data, researchers found that it successfully located the mandibular canals with high precision, only falling short in a mere 1-4% of the analyzed cases. Sahlsten remarks with confidence that this 96% success rate demonstrates the model’s clinical usability and effectiveness.

The unwavering efficiency and speed of artificial intelligence grant it an advantage over human counterparts. This AI model not only expedites the discovery of the mandibular canal but also bolsters radiologists’ and physicians’ decision-making processes. Nevertheless, final treatment decisions ultimately remain in the hands of human healthcare professionals.

In the realm of Planmeca, Finnish family-owned health technology titan, collaboration with CAI and Tays holds immense business potential. The company’s foundation is rooted in 3D imaging devices for dental care and their supporting software. By integrating the neural network model developed during this research into their software, Varjonen believes that they can enhance the usability and performance of their equipment.

Scientific publications produced through this collaboration hold great significance for all involved parties. Some of their findings have even been published in Scientific Reports, a reputable journal. These peer-reviewed publications act as irrefutable proof of the model’s functionality while supporting the thesis work of doctoral candidates.

Equally important to the researchers, these publications aid in the process of obtaining medical device approval for their dental software. According to Varjonen, the publications validate that the software has been designed according to necessary processes and scrutinized meticulously through all required phases.

Beyond locating the lower jaw nerve canal, the collaborative venture between Planmeca, CAI, and Tays also encompassed developing a neural network model for orthognathic surgery, in which facial anomalies in the lower facial region are corrected surgically. Varjonen reveals that this model aids in identifying landmarks in the skull area for planning malocclusion correction and jaw alignment surgery, utilizing the same patient data for yet another AI-driven medical application.

In the foreseeable future, artificial intelligence is anticipated to play a significant role in health applications. Sahlsten envisions AI as an exceedingly potent tool, providing healthcare experts with preliminary assessments or alternative opinions. However, he concedes that deep learning models still face challenges in explaining why they reach specific outcomes, calling for further research to enhance understandability and transparency.

As the cyberpunk future swiftly becomes our present reality, the incorporation of artificial intelligence into various aspects of our lives is inevitable. The alliance of Planmeca, CAI, and Tays exemplifies the power of collaboration and innovation in the progressive field of healthcare.