Using machine learning in the examination of dental X-rays to detect early tooth decay

ISRCTN ISRCTN45982629
DOI https://doi.org/10.1186/ISRCTN45982629
IRAS number 248306
Secondary identifying numbers 18/NI/0111, IRAS 248306
Submission date
27/02/2020
Registration date
03/03/2020
Last edited
15/02/2021
Recruitment status
No longer recruiting
Overall study status
Completed
Condition category
Oral Health
Prospectively registered
Protocol
Statistical analysis plan
Results
Individual participant data

Plain English summary of protocol

Background and study aims
The Manchester Imaging Limited software programme is designed to improve the detection rate of signs of dental disease from x-ray images and therefore diagnose dental diseases earlier than the current standard of care (visual detection by a General Dental Practitioner). This will empower patients to look after their teeth more effectively and potentially save money for the NHS by reducing the need for dental restorations and repairs.

The software will also provide a visual aid for the patient so that they know which of their teeth are under early dental decay or other types of dental disease. This could be more motivational for the patient than the dentist making general statements about frequent brushing and a sugar-free diet.
Previous research indicates that dentists are likely to miss early enamel tooth decay. The current standard care for early enamel decay is the application of fluoride in the form of high concentration toothpaste, varnishes and mouth rinse.

The software needs to be trained using digital images from the most common digital imaging systems used by dentists in the UK. It is thought that the more images that the software sees, the more accurate it will be at detecting dental disease.

The aim of this study is to provide data (in the form of anonymized digital images) by which the sponsor can improve the likelihood of the software programme working to a defined standard across the majority of commercially-available dental digital imaging systems in the UK.

Who can participate?
Patients over the age of 16 attending any of the participating centres where a dental x-ray would be taken as part of standard dental care

What does the study involve?
The study involves dental x-ray(s) images being taken. These will be anonymized for use in training the Manchester Imaging Limited software programme

What are the possible benefits and risks of participating?
There were no direct benefits to the study participants.

Bitewing radiographs (dental x-rays) are part of routine care. For participants taking part in this study, no extra bitewing radiographs will be taken. These procedures use ionizing radiation to form images of teeth and provide a dentist with other clinical information. Ionizing radiation can cause cell damage that may, after many years or decades, turn cancerous. As there is no additional ionizing radiation in this study compared to the normal standard of care, the chances of this happening are the same whether eligible participants take part in the study or not.

Where is the study run from?
9 dental practices in the Northwest of England

When is the study starting and how long is it expected to run for?
July 2018 to January 2020

Who is funding the study?
Manchester Imaging Limited (UK)

Who is the main contact?
Prof Hugh Devlin
hugh.devlin@manchester-imaging.com

Contact information

Prof Hugh Devlin
Scientific

School of Medical Sciences
Faculty of Biology, Medicine & Health
The University of Manchester
Coupland III
Oxford Road
Manchester
M13 9WL
United Kingdom

ORCiD logoORCID ID 0000-0002-5120-747X
Phone +44 01612756849
Email hugh.devlin@manchester-imaging.com

Study information

Study designCross-sectional non-interventional study
Primary study designObservational
Secondary study designCross sectional study
Study setting(s)GP practice
Study typeDiagnostic
Participant information sheet Not available in web format, please use contact details to request a participant information sheet
Scientific titleDental Digital Bitewing Radiograph Study: DDBRS
Study acronymDDBRS
Study objectivesThe software uses machine learning to improve the detection of dental disease. The more images (anonymized digital bitewing radiographs (dental x-rays) from digital scanner systems commercially-available in the UK) that the software learns from, the better it becomes at detecting dental disease.
Ethics approval(s)Approved 12/06/2018, Office for Research Ethics Committees Northern Ireland (ORECNI) (Customer Care & Performance Directorate, Lissue Industrial Estate West, 5 Rathdown Walk, Moira Road, Lisburn, BT28 2RF; +44 028 9536 1400; www.orecni.hscni.net), ref: 18/NI/0111
Health condition(s) or problem(s) studiedDental caries
InterventionThis study will collect anonymized digital bitewing radiographs from the patients of UK general dental practitioners (GDPs/dentists) as part of the normal standard of care to assist this learning process.

Participants were patients attending for their routine dental examination. No additional procedures or activities were carried in addition to the care normally provided to patients at that visit. There was no randomization of participants as there was no control arm to the study.

The Investigator/GDP obtained patient consent during the initial part of the dental consultation. A Patient Information Sheet was given, along with a Patient Consent Form. The potential participant had the opportunity to ask the GDP any questions. This is considered ideal as the GDP is also the Investigator at the study site so was well informed about the study.

An individual’s capacity to consent was determined by their ability to:
• Understand what the taking of radiographs involves, its purpose and nature and why it is being proposed.
• Understand the benefits, risks and alternatives.
• Understand the consequences of not receiving the proposed treatment.
• Retain the information and be able to weigh up the pros and cons in order to arrive at a decision.
• Communicate the decision.

One copy of the Patient Consent Form was kept in the study file, one in the patient record and one handed to the patient. Potential participants were able to consider whether or not to take part in the study during their consultation. As the study involves normal standard of care, the decision whether or not to participate was given at any time during the potential participant's consultation.

Given the nature of the study, the timeframe from consent to the bitewing radiograph being entered into the study is negligible. It is therefore unnecessary to monitor ongoing capacity of the participant.

The patient’s radiographs were then examined for dental disease by the dentists in the normal way and the findings communicated to the patient as part of their normal care. The GDP/Investigator uploaded images/DICOMs to a secure server; these images contain patient identifiable information (PII). Only the Investigator and 7i Group Ltd had access to these DICOMs. 7i Group will anonymize the images (remove any PII) and provide them to the Sponsor. Each image provided to the Sponsor was referenced by a unique ID number. 7i Group maintained a record that linked this unique identifier back to the study site, Investigator and patient. 7i Group destroyed all copies of DICOMs/digital images when the study finished. The Sponsor has kept the anonymized DICOMs.
Intervention typeNot Specified
Primary outcome measureTo receive the required number and required mix of digital images from the majority of manufacturers/models of commercially-available dental digital imaging systems in the UK
Secondary outcome measuresNone
Overall study start date14/04/2018
Completion date17/01/2020

Eligibility

Participant type(s)Health professional
Age groupAdult
SexBoth
Target number of participantsThe theoretical maximum number of study sites is 10 with 150 images from each system. This makes the maximum number of participants 1,500. A participant will be asked for consent at the beginning of their planned dental consultation. During their dental consultation, should the GDP consider that one or more bitewing radiographs (dental x-rays) are necessary as part of normal standard of care, they will be taken; more than one may be taken. NHS Digital data and the Sponsor's previous previous market research suggest that on average, when bitewing radiographs are taken, approximately two are taken.
Total final enrolment1296
Key inclusion criteriaStudy sites:
Use one of the dental digital imaging systems previously established as the most commonly used systems (in market research commissioned by the sponsor) at their dental practice.

Patients :
1. Bitewing radiograph clinically indicated as part of standard care
2. Informed consent to participation given
Key exclusion criteriaPatients:
Aged <16 years
Date of first enrolment27/05/2018
Date of final enrolment17/01/2020

Locations

Countries of recruitment

  • England
  • Scotland
  • United Kingdom

Study participating centres

King Dental & Associates
Council Buildings
Market Street
Carnforth
LA5 9LB
United Kingdom
Hyde Dental Practice Ltd
203 Market Street
Hyde
SK14 1HF
United Kingdom
New Mills Dental Practice
6 Union Road
New Mills
SK22 3ES
United Kingdom
TSa Dental Care
5 Davies Road
West Bridgford
Nottingham
NG2 5JE
United Kingdom
Inverurie Dental Care
22 North Street
Inveruruie
AB51 4QR
United Kingdom
Rudheath Dental Health Centre
144 Middlewich Rd.
Northwich
CW9 7DS
United Kingdom
Aesthetics Dental Solutions
5 Station Square
Lytham
FY8 5PA
United Kingdom
Gaskell Ave Dental
5 Gaskell Ave.
Knutsford
WA16 ODA
United Kingdom
Brunner Court Dental
95 Witton St.
Northwich
CW9 5DR
United Kingdom

Sponsor information

Manchester Imaging Ltd
Industry

Arch 29
North Campus Incubator
Altrincham St.
Manchester
M1 3NL
United Kingdom

Phone +44 01612756849
Email hugh.devlin@manchester-imaging.com
Website https://manchester-imaging.com/about/

Funders

Funder type

Industry

Manchester Imaging Ltd

No information available

Results and Publications

Intention to publish date27/02/2020
Individual participant data (IPD) Intention to shareNo
IPD sharing plan summaryNot expected to be made available
Publication and dissemination planPlanned publication is in a high-impact peer-reviewed journal. We intend to submit work for publication by the end of 2020.
IPD sharing planThe datasets generated during and/or analysed during the current study are not expected to be made available due to the company wishing to maintain its commercial lead and development of intellectual property.

Study outputs

Output type Details Date created Date added Peer reviewed? Patient-facing?
Preprint results non-peer-reviewed results in preprint 14/02/2021 15/02/2021 No No
HRA research summary 28/06/2023 No No

Editorial Notes

15/02/2021: Publication reference added.
02/03/2020: Trial’s existence confirmed by HRA and Health and Care Research Wales (HCRW).