Chronic back pain

Unmet need

Back Pain is a common but poorly understood condition. A small proportion of patient cases are treated surgically.

Magnetic Resonance Imaging (MRI) scanning plays an important part in diagnosis and management decisions. MR spinal imaging is critical to clinical diagnosis and stratification and contributes to outcome assessment.

Spine Registries have been set up to improve patient selection and follow-up outcomes assessment however, up to now, none systematically collect imaging data. Poor stratification (without imaging) is a contributory factor to the limited knowledge of biology of spinal pain and developing rational interventions.

This project will employ Spinenet to assess Spinal MRI images for the need for interventional surgery.

SpineNet is an AI software tool developed in Dept of Engineering Science by a team led by Andrew Zisserman FRS to read lumbar MRI scans.

It was trained on a dataset of 2009 subjects from the Genodisc cohort (PI’s Dr Jill Urban (DPAG) and Prof. Jeremy Fairbank (NDORMS), U. of Oxford).

The images from 8 centres across Europe were annotated by Prof. Iain McCall consultant radiologist. SpineNet was published in 2017, and demonstration annotations are available on SpineNet.

Project aims and objectives

This study is designed to improve our understanding of chronic back pain by developing a new way to extract information from lumbar MRI scans.

Although MRI scans are an important part of diagnosis, many of the changes seen on MRI scans are age related and do not necessarily tell us much about the patient’s diagnosis. This means it is unclear how much the changes we can see on MRI scans relate to backpain.

SpineNet has been applied to Symptomatic and Asymptomatic research cohorts, and the investigators have shown that they can demonstrate the different effects of aging and presence/absence of symptoms on several features in Lumbar spine MRI scans.

The project will recruit BSR (British Spine Register) subjects from OUH and other members of the NCIMI group to provide initial (pilot) analysis.

SpineNet will be used to evaluate how features observed in MRI scans correlate with surgical intervention and the extent to which SpineNet could be used to stratify patients for surgical intervention.

Outcomes

1. To define the contribution of imaging features to a baseline stratification based on the operative intervention.
2. Calculating the probability of a good/poor outcomes for individual procedures based on Imaging features, ODI score, or choice of intervention at baseline