The latest developments in Dental AI can assist dentists to diagnose and treat disease.  However, to make the most of their potential, users need to understand how they work and how they should interpret their output. This talk provided an understanding of how AI works, different ways you can use AI in your surgery and consider if AI in Dentistry has the potential to deliver a real benefit to the profession and patients, or is it just hype and publicity?

Using AssistDent as a case study, Dr Williams spoke about how AI systems detect disease; have the potential to provide increased efficacy in the practice and improve patient care and communication. 

AssistDent is an AI software that detects early interproximal enamel caries in bitewing radiographs. These lesions are easily missed by dentists as they can be difficult to detect due to their small size and image artefacts.  Data of sufficient quality and consistency is required to train AI machine learning, so that the algorithm can learn meaningful patterns.  Data sets also need to be sufficiently large, varied and free from bias to cover all the expected operational input data. He went on to detail some of the AI algorithms and architectures employed by AssistDent and the level of accuracy and automation provided. 

Also welcome and important was the coverage and description of other AI-based dental solutions that are currently available and/or in development and the discussion evolved into how they might provide benefits to dentists and their patients.

This naturally led to crucially acknowledged best practice guidelines for the use of AI by dental professionals, which essentially help improve the clinical accuracy and patient care.

Identification of the common challenges and barriers to adoption, such as integration with the dentist’s existing software and workflow, plus the cost of the equipment and additional training was explored too. 

CPD Learning Outcomes 

·         Awareness of the latest developments in Artificial Intelligence (AI) in Dentistry 

·         An understanding of what AI is and can achieve in dentistry and the potential benefits for dentists and their patients 

·         An understanding of how AI works and best practice in the dental surgery 

·         Awareness of the potential shortcomings of using AI and the challenges of integrating it into a dentist’s clinical workflow 

Dr Ben Atkins owned and ran a group of mixed dental practices in the North West of England.  More recently, he has been actively involved in dental policy including roles such as; President of the Oral Health Foundation, Clinical Advisor to NICE and is participating in various projects with NHS England.

Following Dr Atkins appointment to the Clinical Team as Director of CSR & Regulatory Compliance at Manchester Imaging Limited (specialists in the design, development and deployment of AI systems for the dental sector) he met with Tony Travers CEO and they explored where business and ethical considerations aligned in preventive dentistry and artificial intelligence and where the potential for growth is where the two worlds meet…

Here is the transcript excerpt from their fascinating ‘meeting of minds’:-

“No-one taught me to brush my teeth and my dentist gave me a lolly for being a good patient!”

…reflects Ben, who never missed a six-monthly check-up, yet arrived as a dental student in Sheffield needing eight fillings.  

“The NHS paid for preventive care for me, but did I receive it in a way that I understood, or where I took the responsibility?”

For any busy dental practice and – crucial in the current climate – for a profitable business, a focus on oral health and the transfer of responsibility can be a game-changing approach.

Communication and patient behaviour change are the key drivers as well as the tools to implement such change.  However, the transfer of responsibility requires time and explanation – and dentists, as we know, are notoriously time-poor.  Commitment to emphasising a different philosophical approach can be daunting.  Ben explains how his group embedded this change:

“Essentially, we focused on putting down the drill.  Working to a preventive ethos, we began transferring responsibility to the patient all the while emphasising that this would be a journey.  Our team audited patient lists.  With 20 percent red patients (high needs patients), out of about 140 new patients accepted each month, five to ten patients would be very high needs (Dickensian dentistry at its best).  These high needs patients regularly did not return once they were out of pain, we knew that they could come back if they needed and we had improved their oral hygiene.  

To move to an approach whereby recalls are improved and the opportunity for growth in private practice is enhanced – focusing on both ethical and business imperatives – we found that the critical factor was to introduce a ‘why’?  For me, personally, that was to create time to treat high needs patients.

The aim of this whole journey is to improve oral hygiene and heal decay before we restore, but I must admit that this is possibly the biggest challenge we had when we implemented it in practice. So, we split the treatment, making the first appointment not about individual tooth diagnosis but about the patient’s oral health status.  We were educating patients about the journey we were embarking on together, with a clear message that our team would give them the tools to improve oral hygiene but that they, as patients, would be responsible.  And here, patient education is key. To help achieve this, we developed a patient video and a patient contract, and we spent time going through this contract with the patient – all reinforcing the journey. Patients are responsible for brushing their teeth, we are responsible for education, and one of the biggest learning outcomes for us, as dentists, was the effectiveness and importance of evidence-based advice, which is crucial in communicating the benefits of preventive dentistry to the patient.”

A preventive approach requires commitment and time-constraints can be daunting, seemingly prohibitive.  This is not a straightforward journey, either for dentists or for patients and so new tools and practices that can help mitigate against some of the barriers are invaluable.  Among these, the growth of Artificial Intelligence within dentistry is an exciting development to help facilitate the transition.

Tony comments:  “Our approach at Manchester Imaging is to harness AI’s potential and to raise awareness of its value as an expert colleague within dental practices.  We focused initially on improving the identification of enamel caries in bitewing radiographs on the basis that reversal of enamel-only proximal caries by non-invasive treatments is an important aspect in preventive practice.  Using AI as an additional tool, and as part of a clinical research study, 23 dentists were randomly divided into a control arm without AI assistance, and an experimental arm in which AI assistance provided on-screen prompts, indicating potential enamel-only proximal caries.  All participants analysed a set of 24 bitewings in which an expert panel of dentomaxillofacial radiologists had previously identified 65 enamel-only carious lesions and 241 health proximal surfaces.  The control group found 44% of the caries, whereas the experimental group found 76%, indicating the efficacy of AI as a tool for dentists to enhance the diagnostic capabilities of their radiographs.”

In this instance, the early detection of enamel caries enables them to be reversed with fluoride treatments, improved oral hygiene and dietary advice.  Patients may then avoid later fillings through early preventive measures.  This, in turn, gives dentists more time and greater scope to work within a more balanced practice, providing the opportunity to drive practice growth through prevention.

The patient’s journey to responsibility is not always straightforward and the emphasis on communication cannot be overstated and this is a further aspect of the shift to preventive dentistry where AI can be of great value, as Tony summarises…

“Where AI in general and in this particular instance our AssistDent® tool, can combine both aspects – the need to free up time for dentists and the need to reassure and communicate with patients – is in the rapid detection offered.  Where patients may be hesitant about the preventive concept and the vital importance of taking responsibility, AI tools allow for the immediate presentation of images for discussion between dentist and patient.  This, in essence, is objective evidence, facilitating the precise evidence-based advice that Ben is speaking of when he highlights the need for educating patients.  This can assist in emphasising the nature of the patient’s treatment journey and can act as a clear indication that their preventive regime is working, reinforcing the benefit of the preventive ethos and reassuring the patient of the efficacy of their journey.”

A combined approach – embracing a preventive ethos in practice while harnessing the benefits of AI – can save precious time for dentists; time that can be spent on growing the practice.  

Ben concludes, reflecting on his rationale:  “A preventive approach, bringing in the latest AI tools available, can create a balance within the practice, for dentists and patients alike.  Across our practices, as responsibility moved to the patients, possibly one of the most positive outcomes of this preventive ethos, was the stress it removed from our dentists.  For myself and ‘my why?’ when embarking on the process, it was a time to grow the business and increase its potential and value which enabled me to concentrate my time and energy on high needs patients and contribute more to dental policy.”

Social Media:  Follow Dr Ben Atkins on Instagram: @dentalben

AssistDent® is an aid in the diagnosis of early dental disease and is already used by dentists and in dental schools across the UK and Europe as a prompting system to help when analysing bitewing radiographs, looking for early signs of enamel proximal caries.  

The Innovate UK funding enabled Manchester Imaging, the company behind the AI-based software, to adapt and repurpose their technology to analyse dental radiographs, to deliver a remotely accessible intelligent learning platform for dental students that can record a student’s clinical assessment of a radiograph, automatically interpreting and evaluating it against a gold standard assessment.  

At an early phase of the research process and to evaluate a novel application of AI-based computer aided diagnosis in dentistry training, a pilot comparative study was undertaken, with 24 third-year dental student volunteers randomly divided into control and experimental arms. Both arms examined the same images using the same graphical user interface, with the AssistDent caries detection function disabled for the control arm.

Results of the pilot study demonstrated an increased ability in the detection of enamel-only proximal caries by the students using AssistDent, showing a mean sensitivity level of 0.80 (95%CI ± 0.04), increased from 0.50 (95%CI ± 0.13) p<0.01 shown by students not using AssistDent.  This improvement in ability was achieved without an increase in false positives.

Mean false positives per bitewing radiograph recorded by students when using AssistDent was 2.64 (95%CI ± 0.57), and by students without using AssistDent was 2.46 (95%CI ± 1.51).

A set of example radiographs with expert annotations is available through the platform as a gold standard for student use, or tutors can provide their own images and annotations.

Within the pilot study, gold standard annotation of caries was obtained from a panel of five dentomaxillofacial radiologists and one Professor of Restorative Dentistry, each of whom performed clinical evaluation on a set of images and provided annotation on the location and grade of caries, resulting in a gold standard set of 1,972 examples of enamel-only proximal caries for algorithm training and evaluation.  

The graphical user interface enables students to add their clinical assessment, marking up identified pathologies.  A machine learning algorithm automatically analyses the image and intelligently compares the tutor’s gold standard analysis with that of the student, taking the underlying mouth anatomy into account and giving a detailed breakdown of all the pathology identified (true positives), as well as missed (false negatives) and incorrectly classified (false positives), providing the tutor with a score for each image.  Students can now get a detailed breakdown of their performance and learning progress, in conjunction with their tutors, without needing to physically attend the dental school.

The project to adapt Manchester Imaging’s existing technology was funded through Innovate UK’s “Business-led innovation in response to global disruption” competition – a rapid response funding programme which received more applications than all their competitions for the previous year, combined.  

Neil Morgan, Head of Fast Start at Innovate UK said:

“The Covid-19 situation is not just a health emergency, but also one that affects the economy and society. With that in mind, Innovate UK launched this Fast Start funding, seeking smart ideas from UK innovators. Like other primary care providers, the dental sector has been particularly impacted, so AssistDent’s technology is particularly timely and useful as we seek to get through and recover from the pandemic.”

The award for AssistDent® is clear evidence of governmental recognition of the need to financially support the development of new technologies across the dental profession, and to help companies innovate for public health.

Tony Travers, CEO at Manchester Imaging, comments:

“We are delighted to have received this support from Innovate UK, which has helped facilitate the early and rapid adaptation of our technology to help address current pressing dental educational needs during this pandemic. The sector has been particularly hard hit by COVID-19, both functionally and economically, and we hope the technology now available will continue to assist both institutions and students, harnessing intelligent automation and increasing opportunities for online, remote clinical learning.”

For more information about AssistDent® and this project, please contact:

Tony Travers, CEO, Manchester Imaging Limited

E: [email protected]

M: +44 7776 481645

W: www.manchester-imaging.com

See below:-

NEW:  Risk Analysis for Caries Grading and Tooth Identification Features

NEW:  Improved the performance of the tooth detector

NEW:  Improved caries detection performance

NEW:  Users can edit displayed caries grading and tooth labelling

NEW:  Displayed dentine caries are configurable

NEW:  Introduction of caries grade labels; E1, E2 & D

Improved Caries Detection Performance. Improved caries detection resulting in even more caries being detected, without an increase in the number of false positives.

New Feature: Caries Grading. Extended proximal caries detection to all grades of caries and classified each region of interest as:-

• E1: caries penetrated less than 50% of enamel
• E2 caries penetrated >50% of enamel but not dentine
• D: caries penetrated through enamel into dentine.

Users can edit grading and provide their own grade to added caries regions of interest.

New Feature: Tooth Labelling. Automatic labelling of every tooth according to mouth quadrant and tooth number.  Users can edit and/or correct tooth labels as required.

Improved Tooth Detection Performance. Improved detection of teeth resulting in fewer teeth missed and hence more caries found. Reduced the number of erroneous tooth detections resulting in fewer incorrect caries detections.

Additional Licenses at no Extra Cost. Enabled use on up to 4 computers on a single licence registration to facilitate use across all surgeries in a practice.

Conforming to post-Brexit UKCA mark meaning that AssistDent can be used as a medical device in England, Scotland and Wales.