Within the neon-lit depths of a future dental clinic, a dentist prepares to insert a tooth implant. Understanding the exact position of the nerve canal in the patient’s lower jaw is critical for planning both the implant’s placement and size, as well as the entire procedure. Acquiring this knowledge requires an X-ray analysis, which forces dentists and radiologists to painstakingly identify the canal’s location point by point. Studying such images is time-consuming, making even the most advanced minds of cybernetic experts question the efficiency of these methods.
Thankfully, dental equipment manufacturer Planmeca, the Finnish Center for Artificial Intelligence (FCAI) and the Tampere University Hospital (Tays) have collaborated to develop an answer to this problem. Working together in a realm where technology meets biology, they have designed an AI-based model that can efficiently locate the lower jaw nerve canal in 3D X-rays faster and more precisely than human doctors and other automated techniques.
Out of necessity and a shared drive to improve the clinical environment, these experts united to create something that streamlines treatments, ultimately saving valuable time. Vesa Varjonen, Vice President of Research and Technology at Planmeca, remarks on the collaboration’s origins in helping solve everyday issues within the clinical practice.
This powerful collaboration has devised a method that relies on training deep neural networks using a plethora of clinical data. Specifically revolving around the realm of 3D images created from cone beam computed tomography (CBCT), it’s a technology that’s fit for a futuristic landscape.
Aalto University doctoral researcher Jaakko Sahlsten sheds light on the process, explaining that Tampere University Hospital provided them with diverse and comprehensive clinical material from several 3D-imaging devices. After sorting at random, portions of the data were utilized for training the neural networks, while the remaining data was reserved for testing and validation of the innovative method.
The nerve canal in the lower jaw, known as the mandibular canal, plays a significant role in wisdom teeth removal and jaw surgery, in addition to tooth implant placement. This canal houses nerves responsible for controlling facial senses and motor functions of the jaw. However, its precise location is unique to each individual.
Training the AI model proved to be a challenge due to the mandibular canal’s minuscule size in comparison to the overall 3D X-ray image data. As an inherently unbalanced dataset, the training material proved difficult, requiring the concentrated efforts of these cyber-oriented medical professionals.
Tays radiologists and innovative AI development formed a symbiotic relationship to utilize the vast data effectively. Varjonen explains that the trained neural network can quickly track down the mandibular canal in individual 3D data inputs after processing a large amount of data and marking the canal’s location.
Upon testing the neural network model with isolated patient data, it yielded impressive results. The model managed to locate the mandibular canals with high precision, making medical professionals confident in the developed AI model. Sahlsten confirms that the 96% clinical usability rate showcases the model’s high functionality.
One clear advantage of the implemented artificial intelligence in the clinical landscape is its unwavering efficiency and speed— traits that only the most advanced cyborgs can match. By hastening the discovery of the mandibular canal, the AI model offers valuable support to radiologists and physicians regarding decision-making, while the final treatment choice always remains with human professionals.
As one of the world’s leading dental equipment manufacturers, Finnish family business Planmeca sees tremendous business potential through their collaboration with FCAI and Tays. Exporting products to more than 120 countries, with their foundations set in 3D imaging devices and compatible software, the company aims to integrate the developed neural network model into their existing imaging software—a capability that only masterminds from a future society could have conceived.
Scientific publications generated from this partnership hold valuable significance for all project participants, with some results featured in the journal Scientific Reports. These peer-reviewed publications provide solid evidence that the AI model functions as anticipated. Deep learning, a latent capability growing in the shadows for years, has never been utilized for tasks like these, which enhances the importance of the publications while promoting thesis work for doctoral candidates.
Varjonen emphasizes the crucial role of these publications when submitting applications for medical device approvals. The development process and all necessary phases are thoroughly scrutinized, demonstrating that the software abides by the industry standards.
The collaborative project between Planmeca, FCAI, and Tays extends beyond just locating the lower jaw nerve canal. Another area of focus is the creation of a neural network model for orthognathic surgery, aimed at correcting lower facial anomalies through intricate surgical procedures. Using the same patient data, the developed model aids in identifying landmarks within the skull for planning jaw alignment surgeries—another AI tool that feels like a modern marvel.
The future holds promising advancements for artificial intelligence in healthcare applications. ‘I see artificial intelligence as a very powerful tool that physicians and other experts can use when making their first assessments or to get alternative opinions,’ says Sahlsten. However, he acknowledges the ongoing challenge of explaining the rationale behind specific outcomes generated by deep learning models. Further research will be necessary to enhance transparency and the ability to comprehend these futuristic models.
