13714814 Functional Brain Imaging | Neuroimaging | Functional ...
87918257 Teori Eksistensial Humanistik · 21571704 Parkinson s Disease Wikipedia the Free Encyclopedia · 51395947-Teori-P...
Functional Brain Imaging 25102007 Chimed Jansen The greatest pitfall in neuroimaging is what you don’t see. Given the postulate, the greatest pitfall in neuroimaging is what you don’t see. Three terms require closer inspection. First of all what would be a great pitfall in neuroimaging? Missing a key structural feature in diagnosis, misinterpreting data generated by the system, not being able to monitor a biological process of interest and not being able to understand the relevance of data which is generated are all good candidates. Moving on to neuroimaging, it is clear that there is not a single entity that is neuroimaging, but rather neuroimaging is made up of a collection of techniques developed to examine structural and functional features of the brain. The methods that will be discussed here are CT and MRI on the structural side and SPECT, PET, fMRI, EEG and MEG on the functional side. Finally what is it that we don’t see in neuroimaging? Each of the techniques mentioned has a narrow band of application within which the systems have further restrictions at every level of image generation, with temporal resolution, spacial resolution and sensitivity playing the largest roles. These are not only technical issues waiting to be resolved. There are some basic limitations inherent in each imaging technique. Together these limitations determine what you can’t see. What you don’t see has further implications however, because it includes the things we don’t know we can’t see, or in the words of Donald Rumsfeld; As we know, there are known knowns. There are things we know we know. We also know there are known unknowns. That is to say we know there are some things we do not know. But there are also unknown unknowns. The ones we don't know we don't know. Donald Dumsfeld (US secretary of defence, 2002) It is the unknown unknowns that can lead to pitfalls. These would be for instance, the accuracy of a pharmacokinetic model used in reconstructing a PET image, or the true relation between BOLD and neural activation in MRI. Certain assumptions are made during image reconstruction and interpretation, which may be believed to true, without adequate critical reasoning. A further consideration when discussing what we don’t see, is what is it we are looking for? If our goals are beyond the capacity of the system we are using, we run the risk of interpreting data to match those goals. This is especially clear in the case of the BOLD signal mentioned above. If we
want to see neural signalling, it would be easy to be seduced into overemphasising the connection between the BOLD and energy consumption for neural firing. There is a great deal that we don’t see in neuroimaging, but the extent to which this can lead to pitfalls is specific to the system being analysed. So let us examine each functional imaging technique in detail. First off there are the structural imaging techniques. The goal in CT and MRI is simply to provide the best anatomical image possible with a high contrast between tissue types. In CT this is provided by passing x‐rays through the subject and detecting which areas absorb the most radiation. This is primarily dependant on the calcium level within a tissue. This makes CT ideal for imaging bones, and muscles, and blood vessels. In MRI, 1H atoms are detected, these are primarily in blood, soft tissue and fat. This gives MRI a superior sensitivity in tissue studies.
Figure 1: CT image of a brain. High bone sensitivity. Figure 2: MRI image of a brain. High tissue sensitivity.
These imaging methods are robust and require the least interpretation to be understood. This means the pitfalls in these systems are centred around the spacial resolution and the sensitivity of the systems. Understanding the biological nature of a tissue could be another potential pitfall. For instance, distinguishing between different types of tumours. This is a problem which can be resolved by combining structural imaging with the appropriate functional imaging technique, should it exist. The radioactive tracer based systems, SPECT and PET, provide images through the decay (SPECT) or positron emission (PET) from radioactive isotopes. These systems are commonly combined with CT scanners though combination with MRI is not unknown. There is a lot of potential for diversity in the types of experiments carried out with these systems, as the radioactive isotopes can be incorporated into a range of tracer compounds. An commonly used example is [18F]FDG, an analogue of glucose, which is used determine energy consumption in tissue. The measurements carried out show the distribution of the radioactive isotope through the subject’s body. However determining the meaning of that distribution is less straight than forward. There are a range of confounding factors to take into account. Ideally the tracer should be bound to the desired target, localise in an area of interest and be absent from all other areas. In fact the tracer
molecules distribute throughout the body to some extent, and collect not only in areas which bind with them, but also in areas which store them, such as fat, metabolise them, such as the liver, or excrete them, such as the bladder. Once metabolised their distribution may no longer represent the their binding affinity to the target of interest. Essentially the metabolic process generates an unknown unknown, as the resulting image will not tell us if the model we are using to estimate values for bound ligand, nonspecific binding, metabolised ligand or free ligand, is correct or not.
Figure 3: FDG PET brain scan. Shows energy consumption. Figure 4: fMRI over a T1 MRI image. Shows oxygen exchange.
Moving to fMRI the methodology is quite different, instead of tracers the system measures the blood‐oxygen‐level dependant contrast; that is the amount of oxygenated blood relative to the amount of deoxygenated blood. This value is a measure of oxygen consumption in a specific tissue, however there is an over compensation of oxygenated blood for the amount of oxygen used. So even at this most basic level the system is hard to quantify. The problems only get worse when we try to extrapolate from this data to provide secondary correlates. This is the unkown unknown of fMRI, when we look at an image it won’t show us whether what we see is related to the tissue activity we assume to be responsible, or some other activity. The opportunity for pitfalls is exacerbated again by the fact that most fMRI studies are subtractive so most of the data will not be visible in the final images. The temporal resolution of around 1s, compared to minutes in PET and SPECT, brings fMRI within the outer range of brain activity. This makes the linking the BOLD signal with brain activity highly desirable, because the fMRI could be providing neural correlates. On its own the fMRI can only provide a rough estimate of neural activity, however the correlation can be improved by combining MRI and EEG in a single reading. The EEG can provide a real time guide to neural activity and can be used to guide the data collection and interpretation from the fMRI. EEG and MEG are methods which don’t map the brain in a spacial sense the way the other methods do. Instead they measure the electromagnetic fields generated by bundles of parallel neurons firing in synchrony. This results in a set of wave patterns with varying degrees of regularity. EEG measures
the eclectic field which is strong, diffuse and strongly affected by the skull, so although the temporal resolution is very high, the spacial resolution is very low. MEG on the other hand measures the generated magnetic field which is weak, highly localised and passes undistorted through the skull. This allows MEG to provide a high spacial resolution for activity within three centimetres from the scalp. On their own the EEG and MEG sensors can only provide details on the types of neural firing taking place in the brain and to some extent the intensity of that firing. Using multiple channels allows comparative analysis to identify correlations or the synchronisation likelihood of specific events. MEG can provide detailed maps of networked events between various brain areas. In EEG this is spatially not possible, without combining it with fMRI. Understanding networking within the brain is vital, because if we only look at activation of brain areas we are only observing the source of the signals. This doesn’t tell us anything about the signal pathway related to a cognitive event. EEG/fMRI and MEG are particularly important for the future of the study of higher brain functions, which require interaction between multiple brain areas. Diagnosis of cognitive disorders continues to be greatly improved through developments in this area. However the study of consciousness itself through these techniques is still a long way off. Even if we develop the tools to follow the activity of the brain in real time, will we be able to understand its language? Experience with DNA has taught us that nature can use complex systems to store information. It is highly probable that the method of information storage and transfer in the brain is equally abstract and complex. So is what you don’t see the greatest pitfall in neuroimaging? Well it depends what you are looking for and your understanding of the system in question. If you want a tool for anatomical diagnosis, there are few pitfall unless tissue differentiation or the resolution is inadequate. This requires the least data interpretation and, outside of anatomical understanding, little training is required to avoid these pitfalls. If you want a tool to analyse functionality of specific brain regions in a diagnostic manner, the pitfalls of resolution and indirect measurement tools remain a problem. With some common assumptions being so entrenched as to be taken for granted. Here a greater degree of critical understanding of the system is required to avoid pitfalls. If on the other hand you wish to explore the functionality of brain’s neural networks, the technology is far less precise. The language being used by the brain is only heard as a cacophony or as distant echoes with current imaging techniques and is still far from being translated, so the path is rife with pitfalls for unwary researchers. Image sources: Figure 1: http://elementseven.co.uk/headct.jpg Figure 2: http://www.mtscottimaging.com/uploads/MRIBRAINWEBSITE.jpg Figure 3: http://www.c2i2.org/summer2004/images/Schulthness‐C2I2‐2.gif Figure 4: http://www.itksnap.org/~paul/prj/cmreps/images/fmri_slice.png