The Industry Impact of AI Integration in the Dental Laboratory

This paper explores the ways that artificial intelligence will integrate with dental laboratories and improve computer-aided design/computer-aided manufacturing restorations through accuracy and precision.

In 2021, artificial intelligence (AI) is still only just entering into aspects of various dental workflows. Although we are starting with simpler workflows—mostly related to image identification and crown design—we are looking forward to AI’s expanding influence. Before that future arrives, however, we will need to obtain more data and develop better strategies for how best to use it. The good news is that if we’re intentional with our strategy, we can use the data we already have to start adding more robust AI capabilities to those we have started employing today.

Indeed, AI is poised to drastically change the major areas in which laboratories impact the dental workflow, including diagnostics, design, material choices, and restorative outcomes. Although the influence of AI on these areas has huge implications for the industry and patient experience, the biggest changes will come once we have built up larger databases of patient information from which we can extrapolate accurate, predictive solutions for restorative dentistry.

To get those solutions, however, patient data sharing will need to include information from the wider medical community. Today, dentistry is siloed. Dentists and physicians rarely, if ever, share information with each other. This is an unfortunate hindrance to the frequently touted promise of validating presumptive oral–systemic health links, which should inspire greater preventive dentistry and fewer patients in hospital beds down the road. It is also unfortunate because there is an intimate relationship between the quality and longevity of the restorations we produce and the biological characteristics of the patients for whom those restorations are fabricated. We can leverage AI to gain oral–systemic health insights at the same time that we leverage it to consistently achieve the precision required for the fabrication of restorations with long-term integrity and proper function.

Applying forward-thinking AI development and data sharing strategies to realize those contributions, however, requires that we at dental laboratories—and within the dental industry as a whole—temper our inclination to devote our entire focus to conquering the challenges we face at present. We must divert our resources away from developing AI systems that provide the basic quality controls and efficiencies that seem most useful in laboratories today and direct them, instead, toward a collaborative effort to nurture the yet-to-be-seen potential of more potent AI systems.

Computers are terrific when it comes to spotting the tiniest discrepancies, remembering them and, with the right algorithmic protocols, uncovering patterns in them.

This is not to say that we are misapplying AI today. Those highly specific machine learning solutions currently in deployment should give considerable confidence to anyone with aspirations for a 0% remake rate in laboratory restoration fabrication. In fact, reducing the frequency of remakes was the primary impetus for introducing AI into the laboratory workflow. Although the current 5% industry standard rate is remarkably good given the number of variables at play in restoration production, that 5% hits at the cost center of the laboratory enterprise––an expense that gets passed along to dentists and, in turn, patients and which is compounded as chair time wasted and patient trust spoiled. The current remake rate is as low as it is today because laboratories have made tremendous strides over the past 2 decades thanks to advances in digital technologies and materials sciences, but closing the gap that remains from 5% to 0% remakes will require incremental improvements across a range of subprocesses.

We will find these gains through understanding all the factors that created a remake necessity, which will enable us to predict the probability of certain patients returning, whether due to the complexity of restorative design, the quality of the fabrication or materials, or the clarity of the intraoral scan or physical mold. Of course, human perception and processing is not sufficient to accomplish either of those things. Computers, however, are terrific when it comes to spotting the tiniest discrepancies, remembering them and, with the right algorithmic protocols, uncovering patterns in them.

Most of the AI systems that have entered dental laboratories are, at their core, predictive devices developed to mitigate remake frequency based on their understanding of past failures. When laboratories started looking into machine learning and AI applications 5 years ago, the major motivation was adding a layer of automation to crown and bridge design. Although on the surface automation efforts seem more about manufacturing efficiency than error reduction (the dreaded machine-replaces-laborer scenario), the machine is only able to attain human-level ability by studying human successes and failures. In a sense, error reduction is the primary operating directive for any AI automation system. And because machine intelligence is rooted, fundamentally, in data, it delivers empirically formulated results and does so consistently, unadulterated by habit, bias, fatigue, or any other human condition.

Examles of generative adversarial networks (GANs), 
a type of machine learning algorithm