Purpose

Survival of patients with major burns has improved significantly in the last few decades, but recent studies show that quality of life (QoL) is measurably reduced in burn survivors¹. One year after injury about a fifth of patients meet the post-traumatic stress disorder diagnostic criteria². It is hypothesised that this is caused by neuroinflammatory response following an excessive immune response to severe burns to peripheral tissue³. In this preliminary study we aim to investigate neurophysiological changes following major burns injury to the body and admission to intensive care using high resolution T1 weighted images, resting-state fMRI (rs-fMRI), diffusion tensor imaging (DTI), MR spectroscopy (MRS) and susceptibility weighted imaging (SWI), in a cohort of patients with a broad cognitive deficit.

Methods

15 patients (11 male, 4 female; 25-74 years, mean 40.9 years) and 15 age and sex-matched controls were recruited to this prospective clinical study (1 patient-control pair was excluded from all but rs-fMRI analysis). Patients included were all admitted to the burn intensive care unit at Chelsea and Westminster Hospital between 2009 and 2015 with at least 15% total burn surface area (TBSA). Patients with toxic epidermal necrolysis syndrome, evidence of head trauma, and those for whom MRI was contraindicated were excluded. The study was approved by the research ethics committee and all participants gave informed consent. Cognitive and psychological testing showed patients had lower QoL and a broad cognitive deficit, which correlated with TBSA for certain tests. Participants underwent MRI, including T1 weighted imaging (1mm isotropic voxels), DTI (b=1000,30 directions, 2.5mm isotropic voxels), and single-voxel MRS (15 x 15 x 15 mm) in the posterior cingulate gyrus and the right frontal white matter. MELODIC in FSL software was used to decompose the rs-fMRI data into brain networks using independent component analysis (ICA) and connectivity in these networks patients and controls was compared using dual regression. FreeSurfer was used to extract brain regions from T1 images and compare the normalised volume of the hippocampus, amygdala and precuneus for patients and controls. DTI data were analysed using tract based spatial statistics (TBSS) in FSL and using FreeSurfer regions for the hippocampus, amygdala and precuneus. Cho/Cr, Cho/NAA and Ins/Cr ratios were extracted from MRS data using Tarquin. SWI images were reviewed for the presence of microbleeds by two neuroradiologists.

Results

ICA of all 30 subjects resulted in 17 independent components. Dual regression showed no significant difference in patients compared to controls in 15 of the extracted brain networks, but small clusters of voxels showing a statistically significant increase in connectivity in two networks (Figure 1). When compared to the activation networks defined by Smith et al⁴ one of these ICAs most closely correlated to the “executive control” network and the other to the “default mode network” (cc=0.38 and 0.25 respectively). There were no significant differences for hippocampal and amygdala volumes between patients and controls. TBSS did not yield any statistically significant difference for differences between the two groups. There was also no significant difference for FA measurements in the hippocampus, amygdala and precuneus between patients and controls. Furthermore, no significant difference was observed between patients and controls for Cho/Cr, Cho/NAA and Ins/Cr ratios in both right frontal white matter and the posterior cingulate gyrus. A large number of microbleeds were identified in one patient in the corpus callosum (Figure 2). Two patients had one microbleed while no microbleeds were identified in 11 patients.

Discussion

Severe burn injuries result in a systemic inflammatory response that can cause neuroinflammation; MRI during the acute and subacute phase has the potential to detect neuroinflammation non-invasively and provide prognostic biomarkers of future neurocognitive deficit and provide objective measures of the resultant long term damage. Cognitive testing in this cohort showed deficits covering a broad range of neurological functions. Therefore, it is difficult to identify a particular resting-state or white matter network to target for MR investigation and the pertinence of the small increase in connectivity observed in patients in the rs-fMRI data is unclear. Furthermore, no difference was observed between patients and controls in the other imaging methods used here. It is likely that the low patient numbers, wide range of TBSA (20 to 85 %, mean 38.4%) and particularly the large range in time since injury (9 months to 6.7 years, mean 3.5 years) may have confounded the results. Further study of imaging biomarkers in the early phases of burn injuries is warranted.

Acknowledgements

MEF and MGS contributed equally to this work. MGS is funded by the Brain Tumour Charity and the Brain Tumour Research Campaign