Computational AnalySiS of functionAl Neuroimaging in DPN
Research type
Research Study
Full title
Computational AnalySiS of functionAl Neuroimaging and Drug Response in pAinful Diabetic Neuropathy (CASSANDRA-DN)
IRAS ID
278943
Contact name
Dinesh Selvarajah
Contact email
Sponsor organisation
Sheffield Teaching Hospitals NHS Foundation Trust
Duration of Study in the UK
1 years, 0 months, 0 days
Research summary
Research Summary
One in 16 people in the UK has diabetes and half of these develop nerve damage. This can cause severe pain in the feet and legs. Unfortunately, current medications provide only partial benefit in some, with many enduring inadequate pain relief. Part of the problem is that a ‘one-size fits all’ ‘trial and error’ approach is used hoping to achieve meaningful pain relief. Current treatments at best achieve partial pain relief in only one in three patients. Hence, a new, more personalised approach is needed where the right treatment is given to the right patient first time. Patients with painful diabetic nerve damage (DPN) can be broadly divided into two main groups. The first group comprises of patients with numbness (dead feeling) in the feet yet who have nerve pain in their feet (painful insensate group) and the second group of patients have feet which are sensitive to light touch and/or mild heat (painful sensate group).
More recent studies have suggested that some treatments work better in one group compared to the other. Hence, if we can reliably determine which group individual patients belongs to, we might then be able to offer them the right treatment for their pain. Over the last 10 years, we have performed specialised brain imaging studies known as functional magnetic resonance imaging (fMRI) in patients with painful DPN and identified a number of fMRI measures that can tell these two groups apart. The main objective of this application is to determine if and how fMRI measures can be used to predict the treatment response of an individual patient. We will recruit 40 patients who had previously received lidocaine treatment, 20 whom had responded to treatment and 20 who hadn’t. We will use this group of patients to examine if fMRI can be used to identify patients who had responded to treatment.
Summary of Results
Approximately half of people in the UK with Diabetes develop nerve damage which can cause severe pain in the feet and legs. Current treatments only provide partial benefit in 1/3 patients and it's impossible to know which patients will benefit until after trialling the treatment. Patients with painful diabetic nerve (DPN) damage can typically be divided into 2 groups. The first group comprises of patients with numbness in the feet yet who have nerve pain in their feet (irritable) and the second group of patients have feet which are sensitive to light touch and/or mild heat (non-irritable.) More recent studies have suggested that some treatments work better in one group compared to the other. If we can reliably determine which group each patient belongs to then we might be able to offer the right treatment straight away.
Many patients with severe painful DPN are offered intravenous lidocaine treatment which can provide pain relief for up to 10 weeks in around 40% of patients, who will continue to receive treatment every 8 weeks.
Over the last 10 years, we have performed specialised brain imaging studies known as functional magnetic resonance imaging (fMRI) in patients with painful DPN and identified a number of fMRI measures that can tell these two groups apart.
The main objective of this project is to determine if and how fMRI measures can be used to predict the treatment response of an individual patient.Our first aim was to investigate whether different clinical pain characteristics can be determined by brain connectivity at rest. We began by analysing our existing fMRI datasets, using image analysis tools to determine a unique pain signature for patients with painful DPN. We have programmed this pain signature to distinguish between the different DPN groups. This stage demonstrated differences in connectivity in pain processing brain regions between irritable & non-irritable DPN patients.
Our second aim was to develop a machine learning (AI) model to predict treatment response in patients with DPN, based on the fMRI datasets. We recruited 43 patients, who had previously received Lidocaine treatment, 29 who responded to treatment and 14 who hadn't. All patients underwent detailed clinical & nerve assessments to determine their pain sensory profile & they also underwent brain resting-state fMRIs. After pre-processing we entered the results into a deep learning architecture (a method in artificial intelligence that teaches computers to process data in a way that is inspired by the human brain,) which classified the results into different models.
To our knowledge this is the first study utilising deep learning methods to classify treatment response in painful DPN. Our results demonstrated a high classification performance which means this approach could be used to improve DPN treatment efficiency by ensuring patients receive the correct treatment in the first place.
The next step will be to determine if this model can be helpful to a clinician, if included in a treatment decision.
REC name
South East Scotland REC 02
REC reference
20/SS/0046
Date of REC Opinion
27 Apr 2020
REC opinion
Favourable Opinion