Automatic detection of facial measurements using artificial intelligence
| ISRCTN | ISRCTN17529873 |
|---|---|
| DOI | https://doi.org/10.1186/ISRCTN17529873 |
| IRAS number | 319431 |
| Secondary identifying numbers | IRAS 319431 |
- Submission date
- 19/12/2022
- Registration date
- 03/02/2023
- Last edited
- 14/02/2024
- Recruitment status
- No longer recruiting
- Overall study status
- Completed
- Condition category
- Eye Diseases
Plain English summary of protocol
Background and study aims
The aim of this study is to develop an artificial intelligence (AI) algorithm to provide accurate, reliable and rapid measurements of the face over video consultation to aid clinicians. The algorithms will be used to improve the diagnostic accuracy of video assessment of patients, allowing more patients to be seen in the comfort of their own home and reducing the burden on the already overstretched NHS outpatient services. Patients will be empowered to monitor their own conditions using this AI-assisted technology, where serious cases can be automatically prioritised while reassurance is given to those that are stable or improving.
Who can participate?
Adults over 18 years, with or without eyelid abnormality confirmed by an adnexal specialist
What does the study involve?
Participants will undergo two video consultations with the research team. These will both be done during their normal clinic appointment, in different lighting conditions. During the video consultation, the team will ask the participant to look in certain directions and take short 20-second clips, from which stills will be extracted (photographs) that make up that clip. Participants will also have their eyelid and facial measurements taken manually by a clinician as per normal clinic routine in an oculoplastic service. The total duration of observation will thus be a maximum of 1 hour (including the consenting process and the ability for the participant to ask questions) and there is no follow-up.
What are the possible benefits and risks of participating?
There are no anticipated risks associated with taking part in this study. The benefits will be to contribute to research in using artificial intelligence for producing facial measurements from videos.
Where is the study run from?
Moorfields Eye Hospital NHS Foundation Trust (UK)
When is the study starting and how long is it expected to run for?
March 2022 to August 2024
Who is funding the study?
Moorfields Eye Charity (UK)
Who is the main contact?
Swan Kang, swan.kang1@nhs.net
Contact information
Principal Investigator
Moorfields Eye Hospital NHS Foundation Trust
London
EC1V 2PD
United Kingdom
| Phone | +44 20 7253 3411 |
|---|---|
| swan.kang1@nhs.net |
Study information
| Study design | Single centre development of AI algorithm |
|---|---|
| Primary study design | Observational |
| Secondary study design | Development of AI algorithm |
| Study setting(s) | Hospital |
| Study type | Other |
| Participant information sheet | 42924_PIS_19Aug22_V1.pdf |
| Scientific title | FACE AI: automated detection of oculofacial parameters using artificial intelligence |
| Study acronym | FACE AI |
| Study objectives | To develop an AI computer vision algorithm that recognises and analyses standardised facial, eyelid and periorbital parameters in real-time videos of patients affected by abnormal eyelid conditions. |
| Ethics approval(s) |
Approved 11/05/2023, London - West London & GTAC Research Ethics Committee (The Old Chapel, Royal Standard Place, Nottingham, NG1 6FS, United Kingdom; +44 207 1048 007; westlondon.rec@hra.nhs.uk), ref: 23/LO/0355 |
| Health condition(s) or problem(s) studied | Orbital, oculoplastic, and adnexal conditions |
| Intervention | Development, evaluation and validation of an artificial intelligence computer vision product through prospective database collection and analysis. Participants will undergo two video consultations with the research team. These will both be done during their normal clinic appointment, in different lighting conditions. During the video consultation, the team will ask the participant to look in certain directions and take short 20-second clips, from which stills will be extracted (photographs) that make up that clip. Participants will also have their eyelid and facial measurements taken manually by a clinician as per normal clinic routine in an oculoplastic service. The total duration of observation will thus be a maximum of 1 hour (including the consenting process and the ability for the participant to ask questions) and there is no follow-up. To develop an artificial intelligence (AI) computer vision algorithm to recognise and analyse facial, eyelid and periorbital parameters in real-time videos of patients affected by abnormal eyelid conditions, the researchers will prospectively collect datasets of dynamic videos acquired from both healthy volunteers and patients with eyelid position abnormalities from Moorfields Eye Hospital ensuring underrepresented ethnic background individuals and patients with artificial eyes are included. The researchers will annotate frames of the collected dataset by semantic segmentation to train an AI algorithm that can automatically detect facial, eyelid and periorbital parameters. To validate the AI computer vision algorithm, the researchers will use frames of collected dataset images to fine-tune the AI model. They will perform validation in a hold-out test set from pre-annotated data to deliver a trained and optimised AI algorithm. To evaluate the consistency of the AI computer vision product, the researchers will measure the test-retest reproducibility of automated facial, eyelid and periorbital parameters between different videos of the same patient. To evaluate the consistency of the clinician manual measurements, the researchers will measure the consensus of manual facial, eyelid and periorbital parameters between two expert clinicians. To evaluate the reliability of the AI computer vision product, the researchers will compare the agreement of the AI computer vision product with the clinician’s manual measurements. |
| Intervention type | Other |
| Primary outcome measure | AI computer vision product validated by comparing segmentation and parameters extrapolated with clinician segmentation and parameters using Dice coefficient, mean absolute difference, intraclass correlation coefficient and Bland-Altman analysis at a single timepoint |
| Secondary outcome measures | 1. Reliability of manual measurements assessed by two different clinicians using Bland-Altman analysis at a single timepoint 2. AI computer vision product externally validated using intraclass correlation coefficient and Bland-Altman analysis at a single timepoint |
| Overall study start date | 03/03/2022 |
| Completion date | 29/08/2024 |
Eligibility
| Participant type(s) | Patient |
|---|---|
| Age group | Adult |
| Lower age limit | 18 Years |
| Sex | Both |
| Target number of participants | 100 |
| Key inclusion criteria | 1. Adult patient (aged 18 years or over) 2. Ability to provide informed consent 3. Absence or presence of eyelid abnormality confirmed by an adnexal specialist for the normal and patient group, respectively |
| Key exclusion criteria | 1. Aged under 18 years 2. Unable to provide informed consent |
| Date of first enrolment | 01/02/2023 |
| Date of final enrolment | 31/01/2024 |
Locations
Countries of recruitment
- England
- United Kingdom
Study participating centre
London
EC1V 2PD
United Kingdom
Sponsor information
Hospital/treatment centre
162 City Road
London
EC1V 2PD
England
United Kingdom
| Phone | +44 20 7253 3411 |
|---|---|
| jamie.webb6@nhs.net | |
| Website | http://www.moorfields.nhs.uk/ |
"ROR" |
https://ror.org/03zaddr67 |
Funders
Funder type
Charity
Private sector organisation / Research institutes and centers
- Location
- United Kingdom
Results and Publications
| Intention to publish date | 01/11/2025 |
|---|---|
| Individual participant data (IPD) Intention to share | No |
| IPD sharing plan summary | Not expected to be made available |
| Publication and dissemination plan | Planned publication in a high-impact peer-reviewed journal |
| IPD sharing plan | The datasets generated during the study are not expected to be made available due to patient identifiable data and images. Data will not be shared. |
Study outputs
| Output type | Details | Date created | Date added | Peer reviewed? | Patient-facing? |
|---|---|---|---|---|---|
| Participant information sheet | version 1 | 19/08/2022 | 09/01/2023 | No | Yes |
| HRA research summary | 20/09/2023 | No | No |
Additional files
Editorial Notes
14/02/2024: The overall study end date was changed from 29/02/2024 to 29/08/2024.
15/01/2024: The following changes were made to the study record:
1. The recruitment end date was changed from 01/01/2024 to 31/01/2024.
2. The overall study end date was changed from 01/12/2024 to 29/02/2024.
20/09/2023: A link to the HRA research summary was added.
11/07/2023: The following changes were made to the trial record:
1. The recruitment end date was changed from 01/07/2023 to 01/01/2024.
2. The overall end date was changed from 01/06/2024 to 01/12/2024.
3. The intention to publish date was changed from 01/05/2025 to 01/11/2025.
4. The plain English summary was updated to reflect these changes.
5. The ethics approval was added.
09/01/2023: Trial's existence confirmed by Moorfields Eye Charity.
