AN-EYE-4-PAIN
Research type
Research Study
Full title
An EYE for Artificial Intelligence for the Diagnosis of PAINful Diabetic Peripheral Neuropathy
IRAS ID
329117
Contact name
Uazman Alam
Contact email
Sponsor organisation
University of Liverpool
Duration of Study in the UK
0 years, 6 months, days
Research summary
This study aims to expand the scope of a prototype lab-based computer (artificial intelligence) model, which we have developed using images of the small nerve fibres at the front of the eye, to characterise patients with painless and painful DPN. We also aim to develop this model to detect the presence or absence of a process by which pain signals are amplified in the spinal cord. The model will be developed using an existing dataset of images during the initial phase of the study. The pilot study will use data acquired from newly recruited participants to provide preliminary information about the accuracy of the model. There will be no clinical decision or intervention based on the output of the model.
Damage to the nerve fibres in individuals with diabetes (diabetic peripheral neuropathy (DPN)) is the most common cause of nerve damage (neuropathy) worldwide, affecting around half of all people with diabetes (260 million globally). Pain as a result of this nerve damage (neuropathic pain) is also a major feature of DPN in around one-third of all patients. Small nerve fibres (pain generating nerve fibres) are the earliest to be damaged and frequently small nerve fibre loss and neuropathic pain precede abnormalities on clinical examination and the most commonly used diagnostic tests. Therefore, there is an unmet clinical need to accurately detect DPN.
Over the past 20 years, we and others have demonstrated that CCM, a technique that allows us to see and take images of the layer of small nerve fibres in the cornea which are the same type of nerves that are involved in neuropathic pain elsewhere in the body, can detect DPN at an earlier stage than conventional methods. This technique is real−time, non−invasive, time−efficient and repeatable. Additionally, CCM has also shown differences in corneal nerve structure between painless DPN and pDPN, with a smaller number of small nerve fibres seen in individuals with pDPN.
Currently, we do not know whether distinct CCM image characteristics can be used to determine presumed pain generation processes and, therefore, guide the treatment choice that targets these processes. One potentially important pain generating process is spinal disinhibition, whereby a failure to suppress or exaggeration of incoming signals, resulting in amplification of pain. A biological marker of spinal inhibition is the change in the size of a reflex called the H-reflex (HRDD). We have demonstrated that HRDD is impaired in patients with pDPN.
This study is based on our recent novel AI computer modelling. Over the last two years, our group has developed and presented the first self-learning model using CCM. We aim to build upon this technology to produce two further models which differentiate painless DPN versus pDPN and the presence or absence of spinal disinhibition.
The patient and societal benefits from this study are considerable, if the projected outcomes from this study are met, and if CCM predicts those who may benefit from targeted therapy.REC name
London - Westminster Research Ethics Committee
REC reference
23/PR/0629
Date of REC Opinion
10 Jul 2023
REC opinion
Further Information Favourable Opinion