This study describes a wide variety of mostly subtle changes on brain MRI in adults with severe falciparum malaria. None of the changes were more frequent in those with coma compared to severe disease without coma, or in patients with fatal disease compared to survivors. Diffuse mild brain swelling was present in the majority without evidence of diffuse cerebral oedema, either cytotoxic (on DWI) or vasogenic (T2 and FLAIR) in most. This swelling was probably at least partly due to venous congestion of the sequestered parasitized red cell mass causing increased cerebral blood volume, whereas no venous thrombosis was detected. The observed brain swelling was considered insufficient to cause coma and was in addition not specific to coma or fatal disease in this series and not correlated to ONSD. The MRI changes thus appear not to be in a causal relationship with observed neurological symptoms. The DWI abnormalities observed coincided with low ADC and most likely represented cytotoxic oedema, perhaps related to ischaemia, or possibly acute demyelination.
In this study, most patients had retinal whitening which was mostly diffuse small patches and more common and severe in those with coma. The retina is part of the central nervous system (CNS) and whitening is thought to be due to ischaemia as a result of heterogeneous obstruction of the microvasculature by sequestered parasites . As the retinal vasculature is very similar to that in the brain, it strongly suggests that ischaemia would also occur in the brain. In autopsy studies the amount of sequestration and microvascular congestion in the brain has been shown to correlate with coma in malaria . On brain MRI in the present study there was a lack of diffuse ischaemic changes on DWI and complete absence of haemorrhages on GRE. This may be because MRI is insufficiently sensitive to detect small lesions of very focal ischaemia. Retinal lesions seen in this study were typically 0.2-0.5 mm in diameter and brain haemorrhages on post-mortem studies of fatal malaria are typically microhaemorrhages . 1.5T MRI is limited to 1 mm for T1 and T2 and 2 mm or more for DWI . Ultrahigh-field MRI at 7 or 8T would be required to show lesions of this magnitude on T2/FLAIR imaging  and 1 mm on DWI . Such scanners are generally not available in the resource poor areas where malaria is common.
A minority of patients had raised lactate on MRS, although raised CSF lactate in cerebral malaria is common . MRS in this study was limited to a single voxel in the parietal cortex but ischaemic lesions were predominant in the brainstem and basal ganglia. Systemic markers of ischaemia (blood lactate) and sequestered biomass did not correlate with the findings on MRI. This may reflect the heterogeneous distribution of parasite sequestration in different organs in the body as shown in autopsy studies . Explanations for this include differences in endothelial cell surface receptors in different tissues  and between individuals.
The MRI and retinal findings in this study contrast to those seen in previous studies in African children . In both populations, lesions in the brain in severe malaria were found in a broad range of anatomical locations. However, the type and severity of abnormalities seen was markedly different. The most common abnormality in Malawi was basal ganglia lesions, present in >80% compared to 23% in the present study. Brain swelling was much less severe in the present study than in African children [4, 22]. In previous imaging studies most adults with cerebral malaria had little evidence of cerebral oedema [23, 24], or showed mostly moderate brain swelling not correlating with coma depth . High signal on T2 associated with thickening of the supratentorial cortex was present in the majority of Malawian children . In contrast to Bangladeshi adults, this suggests the swelling in Malawian children was at least partly due to oedema. In many of these children, the T2 changes were confluent and in some associated with diffuse abnormalities on DWI. These larger lesions were not seen in adults in the present study and this mirrors differences in the lesions seen in the retina; confluent patches of retinal whitening being common in African children with cerebral malaria but absent in Bangladeshi adults [26, 27].
In adults, a slight increase in brain volume has been attributed to increased intracranial blood volume probably as a consequence of sequestration of parasitized erythrocytes . The present study appears to confirm this by finding brain swelling and venous congestion without signs of increased ICP. Raised intracranial pressure in children  and the extent of brain swelling on CT in adults [25, 29] are unrelated to mortality and depth of coma. Mannitol to reduce ICP in cerebral malaria did not improve outcome in adults  or children . The exact role of raised ICP in the pathogenesis of cerebral malaria is unclear, but seems to play only a minor role in adults. Rather than a primary cause for coma it is more likely a feature developing in the later stages of the disease.
This study had several limitations. It was not possible to perform all MRI sequences in all patients due to limited availability of scanners and patients being too unwell or restless. MRI was not performed in half of the fatal cases who died shortly after admission, which could have confounded the selection of patients. However, as most MRIs were done on the day of admission, and all within 48 hours, it seems unlikely that this will have had a major effect on the findings. The scans were assessed by a single observer. Interobserver variability could thus not be quantified.
The mechanisms of coma and death in malaria are probably multifactorial and individual factors might contribute to different degrees between individuals. Subtle variations in the amount and location of sequestration and swelling may lead to coma in some individuals but not be apparent on MRI. Sequestration could target neurotoxic substances produced by the parasite. Changes in areas that do not determine consciousness could be obvious on MRI but not result in coma. In addition, metabolic disturbance outside the brain may cause coma and death with a normal MRI appearance.
MRI has great potential to further elucidate the pathogenesis of coma and death in malaria. Future studies should use MRS to study metabolic disturbance in different parts of the brain and gadolinium contrast to quantify cerebral perfusion and map venous congestion. The availability of increasingly sophisticated scanning software and more powerful scanners should greatly assist in these efforts.