An endoscopy image database for research and training. Version 2.0
Research type
Research Database
IRAS ID
326599
Contact name
Laurence Lovat
Contact email
Research summary
An endoscopy image database to improve the detection and characterisation of gastrointestinal pathology
REC name
East of England - Cambridge Central Research Ethics Committee
REC reference
23/EE/0106
Date of REC Opinion
19 Jun 2023
REC opinion
Further Information Favourable Opinion
Data collection arrangements
Study number for patient procedure (this is a number generated for the research database only – this cannot be used to re-identify patients)
Patient demographics (such as age by decade, biological sex, ethnicity)
Patient medical data (such as major co-morbidities, anti-coagulants / anti-thrombotic medication, relevant GI conditions)
Procedure performed (month of procedure, colonoscopy Indication, recurrence of a lesion)
Procedure extent (anatomical landmark)
Image processor used
Endoscope used
Total number of lesions imaged
For each lesion imaged (study number generated for each lesion for the research database only):
Image type: Digital still image or video
Lesion Size (mm) and classification using standard classification systems such as Paris and Prague
Location within gastrointestinal tract
Resection information including bleeding / perforation
Imaging modalities used
Optical diagnosis
Histopathology (if applicable)
Image classification
Image type
Image findingsThe non-identifiable image files on the encrypted USB flash disks or hard drive will then be transferred to the research team at the Wellcome/EPSRC centre for Interventional and Surgical Sciences (WEISS) at UCL. The data will be uploaded to a secure UCL server within the computer science department which will host the database.
Alternatively, non-identifiable image or video data can be transferred electronically to a secure electronic database in the same way that non-image data is transferred.Research programme
We propose that an endoscopy imaging database would allow for significant research advances leading to multiple projects in the following areas for example: 1) Development and comparison of methodologies for training, accreditation and quality assurance in endoscopy 2) Development of deep learning algorithms and computer vision techniques to allow for automated computer aided detection and characterisation of lesions and landmarks 3) Development and validation of novel classification systems for characterising lesions and landmarks 4) A comparison in the accuracy of endoscopic optical diagnosis using different imaging modalities 5) Comparisons in endoscopic image interpretation between endoscopists and deep-learning computer software 6) Development of deep learning algorithms and computer vision techniques to allow for automated measurement of existing and new quality metrics in endoscopy 7) Development of computer software using deep-learning techniques to automate endoscopy reporting 8) Creation of a future robust research platform to ensure the above objectives are continuously developed as novel imaging techniques emerge over time The ultimate aim is to use the database to produce high impact research outcomes and training resources leading to an improvement in the quality of endoscopy performed and a reduction in missed cancer rates and associated mortality. No other databases of this nature exist as far as we are aware. This database will be made available for other users in line with access arrangements set out in section 19 of this application. We believe that this will offer a valuable resource to interested groups around the UK.
Research database title
An endoscopy image database to improve the detection and characterisation of gastrointestinal pathology
Establishment organisation
Wellcome/EPSRC Centre for Interventional and Surgical Sciences, University College London
Establishment organisation address
Charles Bell House
43-45 Foley Street, London
London
W1W 7TS