MRI: What is it good for?

We are being constantly bombarded with news stories containing pretty pictures of the brain, with headings such as “Brain’s adventure centre located“. Journalists now seem to refer routinely to functional magnetic resonance imaging (fMRI) as “mind reading”, and exaggerated claims about its powers abound, as do misleading, irresponsible and downright ridiculous stories about the technology.

Take, for example, this article by Jeffrey Goldberg in The New Atlantic:

The preliminary findings began to arrive a few days later, in a series of e-mails…”Carter: big amygdala response on both sides! Jeff, do you fear this guy?” Fear might not be the most accurate term, but I worry about him a great deal. I’d recently given his book on Israel a negative review in The Washington Post. Score one for the fMRI.

Yes, the amygdala is involved in encoding fearful memories, but it is also involved in many other emotional responses, any of which could cause an increase in activity there. Yet for Goldberg, a colourful blob on the amygdala must mean he must be scared, or at least apprehensive of, Carter, when in fact this is not necessarily the case.

fMRI is not mind reading, and never will be. The closest that researchers have got to reading minds is the accurate prediction of which of several visual stimuli is being viewed by a subject. This can be done by focusing on the activity in two discrete regions of the brain. For example, activity in the fusiform face area is known to increase when a face is being viewed, while increased activity in the parahippocampal gyrus is associated with the viewing of images of places, so this kind of prediction can be made by monitoring the relative activity in those areas.

The important thing to remember about such studies is that the number of variables and possible outcomes is extremely limited. Neuroimaging is only just moving toward multivariate ananlysis (looking at the activity of many regions simultaneously), so we’re still a very long way away from any real understanding of how hundreds or thousands of modules act in parallel and in concert to generate behaviours, or of the continual stream of consciousness that we experience every waking second in the real world.

The controversy now surrounding neuroimaging stems in part from the recent hijacking and politicization of fMRI data, vis-a-vis Goldberg, and partly from the claims made by two companies that fMRI can be used to detect deception. Despite the fact that these claims are unfounded, neuroimaging data is now admissable as evidence in American courts of law.

Actually, the interpretation of fMRI data is fraught with difficulties, and nothing is as straight forward as we might be led to think. Moreover, we have very little understanding of how the technique works; two studies published back-to-back just in this past week have given us some insight into how the fMRI signal is generated.

Vaughan provides an excellent summary of the fMRI debate, complete with links to essential reading on the subject; Jake dissects Goldberg’s article; and Dave discusses a recent study which concludes that adding a pretty picture of a brain makes data seem somehow more “scientific” and therefore more credible and convincing. Below, I will discuss some of the more useful applications of fMRI and related techniques, namely, identifying individuals at risk of Alzheimer’s Disease and establishing the relationship between focal pathophysiology and cognitive function in various other neurological conditions.

Following recent advances, fMRI can now be used to identify those at risk of Alzheimer’s Disease. This is possible because we know that the hippocampus is the first part of the brain to begin to degenerate in the condition, and that the rate of hippocampal atrophy accurately predicts cognitive decline.

This has all been established in longitudinal fMRI studies in which each patient is scanned repeatedly over a period of up to 5 years. Volumetric data taken at different times can be compared, using a semi-automated method whereby the major sulci are delineated and then aligned manually. The precise matching and digital subtraction of serial images allows for an accurate assessment of atrophy progression, with each patient acting as their own control.

The largest such study was carried out by Ridha et al in 2006. This was a serial MRI study of progressive atrophy in familial Alzheimer’s Disease: 9 members of the same family, each carrying an autosomal dominant mutation the Amyloid precursor (APP) or Presinilin 1 gene, were compared to 25 healthy controls. Each participant was scanned repeatedly, over a period of nearly 15 years, with each scan producing 124 contiguous 1.5mm-thick coronal slices, which were used to measure whole brain and hippocampal volumes. The carriers had clinical assessments and fMRI scans at different clinical stages of the disease (preclinical, mild cognitive impairment and clinical Alzheimer’s). The progression of atrophy could thus be measured both within each stage and across the transition from one stage to another.

In carriers, differences in hippocampal atrophy rates were evident more than 5 years, and differences in whole brain atrophy more than 3 years, before clinical diagnosis. These differences were not significant until the mild cognitive impairment stage, but as the carriers progressed through the clinical stages of the disease, the rate of both hippocampal and whole brain atrophy accelerated. Structural changes were observed by fMRI before the clinical onset of any symptoms; by the time of diagnosis, the carriers had ~18% smaller hippocampal and ~5.5% smaller whole brain volume than the controls. The rates of hippocampal and whole brain atrophy were found to increase by 3% and 0.5%, respectively, per year in the mild cognitive impairment stage, suggesting that this may represent an early stage of Alzheimer’s.

This study is particularly informative, because the early onset of familial Alzheimer’s meant that there was minimal co-morbidity (e.g. vascular dementia) which mnight otherwise confound the results. However, familial Alzheimer’s is rare, accounting for less than 1% of all cases, and it should not be assumed that the findings can be applied to sporadic cases of Alzheimers, which constitute the vast majority of cases .

Diffusion/ perfusion imaging is very useful in detecting cerebrovascular abnormalities in children with sickle cell disease (SCD). Children with SCD are 250 times more at risk of stroke, and cerebral infarction occurs in up to 20% of cases. This is associated with reduced IQ and deficits in executive function and attention. Perfusion imaging can be used to investigate cerebral bloof flow, cerebral blood volume and mean transmit time in children with SCD.

One study looked at 48 young patients with SCD; 25 of these were found to have perfusion abnormailites, and 24 of these were symptomatic. In those presenting with symptoms, MRI showed that cerebrovascular abnormalities were most often found in the internal carotid artery and the anterior, middle, and posterior cerebral arteries. This was accompanied by reduced blood flow and prolonged mean transmit time.

The same technique also provides a novel approach to investigation of acute stroke. In the first few hours following a stroke, tissue showing diffusion and perfusion abnormalities is likely destined for infarction, whereas tissue with perfusion but not diffusion abnormalities is somewhat compromised, but not enough to lead to infarction. Diffusion/ perfusion imaging can be used to detect the areas in which these mismatch occurs; these areas are potential targets for therapy which can minimize or reduce the irreversible damage that occurs as a result of a stroke.

fMRI has also been very useful in identifying structural abnormalities underlying a speech disorder caused by a mutation in a gene called FoxP2, and in clarifying the role of that gene in speech and language. The disorder, a verbal dyspraxia caused by a defect in orofacial movements, is seen in about half the members of a family known by the initials KE, who have been studied extensively in recent years. It was known to be neurodevelopmental in origin, and proposed to occur due to bilateral disruption of the speech circuitry, because left hemisphere damage alone in young children does not produce severe language deficits because brain plasticity allows for sufficient functional compensation.

Using fMRI, researchers showed that, in comparison to healthy controls, KE family members with the speech disorder had abnormally low grey matter density in the inferior frontal gyrus, which contains Broca’s speech area, and in the precentral gyrus, which contains the primary motor cortex. The caudate nucleus, a part of the basal ganglia known to be involved in movement, was also found to have a 25% reduction in volume. There was, however, a higher density of grey matter in Wernicke’s area, another speech centre found towards the back of the temporal lobe.

Neuroimaging linked these structural abnormalitites to specific cognitive deficits. In one experiment designed to test for semantic retrieval and planning of articulation, the participants were asked to silently generate verbs in response to nouns presented to them. Controls and unaffected KE family members showed increased activity in most of the inferior frontal gyrus; in family members with the speech disorder, there was greater and more diffuse involvement of the right hemisphere, but no increase in activity in the left frontal lobe, including in Broca’s area.

In another experiment, the participants were asked to generate verbs out loud, and to repeat words presented to them. Under these conditions, affected members of the KE family had similar patterns of activity to the healthy controls and their unaffected relatives, but there was significantly, less activation of Broca’s area.

Overall, these showed that those with the FoxP2 mutation did not have a deficit in semantic retrieval per se, but rather a deficit in the rapid selection of appropriate items from semantic memory. There was no compensation by the right hemisphere in carriers, confirming that the FoxP2 mutation causes bilateral abnormalities in the fronto-striatal speech and language network; the results also suggest that the FoxP2 gene plays an important role in the development of this network, which is also involved in learning and/ or planning and execution of the sequences movements required for speech. Thus, members of the KE family carrying the mutation have a developmental verbal dyspraxia which affects articulation but not comprehension of speech. The structural abnormalities detected by fMRI had not previously been observed with conventional neuroradiological assessments.

Functional neuroimaging has also been used to identify a previously unrecognized condition now known as developmental amnesia, which is caused by perinatal hypoxic-ischaemic injury (or reduced blood flow to the brain during birth). This type of injury, normally leads to cererbal palsy, seizures or mental retardation, depending on the extent of injury. But it turns out that in some cases the hypoxia is insufficient to cause noticeable changes in behaviour or cognitive abilities. Children who incur a mild hypoxic injury during birth have severe hippocampal abnormalities that lead to severly impaired episodic (or autobiograpaphical), but not semantic, memory

This condition would not have been recognized were it not for the particularly high sensitivity of one type of fMRI (T2 relaxometry) to hippocampal abnormalities. For the same reason, this technique are also especially useful in the presurgical evaluation of patients with intractable epilepsy, in which the tissue where the seizures originate can be precisely located before surgical resection.

Gadian, D. G. et al. (2000). Developmental amnesia associated with early hypoxic-ischaemic injury. Brain 123: 499-507. [Abstract]

Kirkham F. J., et al (2001). Perfusion magnetic resonance abnormalities in patients with sickle cell disease. Ann Neurol: 49: 477-485. [Abstract]

Liegeois, F. et al (2003). Language fMRI abnormalities associated with FOXP2 gene mutation. Nat. Neurosci. 6: 1230-1237. [Abstract]

Ridha, B. H. et al. (2006). Tracking atrophy progression in familial Alzheimer’s disease: a serial MRI study. Lancet Neurol., 5: 828-834. Abstract

2 thoughts on “MRI: What is it good for?

  1. On the subject of fMRI and its (mis-)interpretation, I’d like a neuroscience person’s opinion of this passage from Leonard Mlodinow’s The Drunkard’s Walk (2008). I found it a good book, overall, worthy of being shelved alongside Innumeracy (1988) and How to Lie with Statistics (1954). However, a (mostly incidental) paragraph in the opening chapter caught my eye; I think it’s a reasonable summary of the journal articles cited, but that it’s easily misread.

    Functional magnetic resonance imaging, for example, shows that risk and reward are assessed by parts of the dopaminergic system, a brain-reward circuit important for motivational and emotional processes.1 The images show, too, that the amygdala, which is also linked to our emotional state, especially fear, is activated when we make decisions couched in uncertainty.2

    The footnotes point to the following papers:
    1. Kerstin Preuschoff, Peter Bossaerts, and Steven R. Quartz, “Neural Differentiation of Expected Reward and Risk in Human Subcortical Structures,” Neuron 51 (August 3, 2006): 381–90.
    2. Benedetto De Martino et al., “Frames, Biases, and Rational Decision-Making in the Human Brain,” Science 313 (August 4, 2006): 684–87.

Comments are closed.