Artificial Intelligence in Dentistry

So what exactly is artificial intelligence? Artificial intelligence is the study and process of training computers to do tasks that, at present, humans can do better. Machine learning (ML), one of the many forms of AI, is the development of computer software algorithms that allow computer programs to automatically improve through experience. Machine learning requires building a comprehensive dataset, labeling and classification of the data within the dataset, and then training the computer software to provide predictable, accurate outcomes. There are several different methods of ML that may involve supervised, unsupervised, or reinforced learning. Regardless of the method, what is common in all of them is the comprehensive dataset used for the learning process.

What most people do not realize is how much artificial intelligence and machine learning (AI/ML) is already a part of our lives. If you use Amazon Alexa, Google Home/Assistant, or Microsoft Cortana, you are using and taking part in AI/ML. The questions you ask and the dialect or accent you speak are used by ML ultimately to improve the performance of the system. AI/ML software is built into recent Google Pixels, iPhones, and Samsung smartphones. Selfies and portrait-mode camera features are continuously improving due to AI/ML software inside the phone.

In dentistry, AI/ML is already being used in dental radiographic analysis and interpretation. Object identification and classification for such things as teeth, tooth number detection, dental implants, and restorations have been demonstrated. Services for automatic detection and charting of pathologic conditions such as apical periodontitis and caries using AI/ML are being offered today.

From a restorative perspective, at least one chairside scanner manufacturer claims to use AI to eliminate unwanted soft tissue gathered during intraoral digital impression scans. More interestingly, AI/ML is also being used successfully to design crowns in a chairside dental CAD software program.

In politics, there is a phrase, "follow the money." In the world of AI/ML the phrase is more appropriately, "follow the data." ML cannot take place without data. The more data available and the better the algorithms, the better the output. But how far that information goes and how it is secured will be important questions raised alongside the rise of this technology.

Where will this go in the dental industry? It depends on the quantity of data, who has access to it, and how much more can continue to be gathered. Crowdsourcing involves continuous data input from a group of users. Waze is a smartphone app that uses crowdsourced traffic data to provide real-time navigation guidance and predictable outcomes based on current and historical data. Could there ever be a crowdsourced equivalent that combines dental radiology, 3D modeling, and/or diagnosis and treatment records from dental professionals with medical health records?

From a clinical perspective, the future is likely to bring not only the detection of additional abnormalities in radiographs, but also the identification and potentially even diagnosis when compared to thousands—if not millions—of other radiographs.

Suppose patients had digital impressions taken annually that could be compared and analyzed to show trending wear patterns and deterioration. Could that information be compared with similar wear trends in other patients to identify and predict future maladies, including those that could be prevented with proactive treatment planning before further deterioration occurs?

It is only a matter of time before AI/ML technology becomes commercially available in mainstream crown-and-bridge design software for dental laboratories. However, the progress is likely to be slow until there is enough data collected, shared, and processed by the ML algorithms and integrated into the CAD software.

For production centers with sufficient volume and manufacturing equipment, it's possible that AI/ML could be used to optimize the CAM software nesting or manufacturing processes. With robotically controlled stain applicators, AI/ML could be used to apply characterizing stains to milled zirconia restorations based on learned stain patterns and case-specific requirements.

Artificial intelligence and machine learning software have also gained the attention of the FDA. The FDA already has high expectations for software used in the design and function of medical devices. Verification and validation of traditional software is already an extremely regimented and complicated process. Validating software which has the ability to adapt and change its output based on learned information can present significant challenges. The FDA is currently trying to work through how this will affect clearance and compliance obligations for medical device manufacturers.

There is no doubt that artificial intelligence and machine learning will have an impact on the dental industry in the coming years. It has great potential to improve patient diagnoses and treatment outcomes. The impact to dental laboratory workflow is a little less certain and maybe even a little ominous. In the meantime, if my dental CAD workstation says to me, "I'm afraid I can't do that, Chris," it may be time to seek another line of work.

Chris Brown, BSEE
Manager of Aclivi Consulting
Pinckney, MI