Identification of a biomarker of sensory dysfunction for patients who have developed neuropathic pain following cancer chemotherapy using functional Magnetic Resonance Imaging (fMRI) and machine learning.

Background
Acute pain is intrinsically unpleasant and aversive but is useful as it has survival benefit. However, many people suffer from chronic treatment-resistant pain that lasts months or years, lacks survival benefit, seriously impairs quality of life and causes unnecessary suffering. Biomarkers have been crucial in many areas of medicine for guiding clinical practice and research. However, despite considerable interest in identifying brain biomarkers, there are significant challenges to identifying a clinically useful pain biomarker (Mouraux and Iannetti, 2018). This study will focus on a common cause of neuropathic pain: cancer chemotherapy induced peripheral neuropathy (CIPN), in particular those patients who develop sensory dysfunction (Han and Smith, 2013).

Aims & Objectives
• High accuracy, sensitivity and specificity of prediction of sensory dysfunction for individual patients with CIPN using machine learning and functional magnetic resonance imaging (fMRI).

Method
Patients will be recruited and clinically assessed by a band 5 research nurse employed part-time 1.5 days/week for 2 years to support the study. Two groups of 20 patients will be identified: i) patients with chronic sensory neuropathy as a prominent feature of post-chemotherapy CIPN and ii) patients who have received chemotherapy but not developed or recovered from neuropathy. Patients with mild-moderate mood and anxiety symptoms will not be excluded; however, patients with serious systemic disease or varying intensity of perceived pain will be excluded. Patients will be characterised using the Brief Pain Inventory, standard mood and anxiety rating scales and Quantitative Sensory Testing (QST) (Scott et al., 2012). A blocked fMRI design will be used with manual brush-evoked allodynia applied according to a previous description for affected and unaffected body sites lasting about 16 min (Schweinhardt et al., 2006). Resting state fMRI will be acquired for 10 min. Machine learning algorithms will then be applied in a within study replication (cross-validation) framework according to previous work in Dundee [see machine learning references] and individual patient accuracy, sensitivity and specificity of the predicted presence of sensory disturbance determined. The student need not participate in patient recruitment and clinical assessment (unless they wish to do this as the nurse will cover these activities).

Student Background
This study would be particularly suited to a student with a quantitative first degree (e.g. physics, mathematics, engineering, computer science), who aims to develop a career in machine learning applied to medicine. Academic and commercial interest in applying machine learning to medicine is very rapidly expanding and this project provides an opportunity to develop expertise in this area. As above, the student need not participate in patient recruitment and clinical assessment (unless they wish to do this as the nurse will cover these activities).

Apply
To apply please send a cover letter, curriculum vitae and two references to: [Email Address Removed]