[music] Dear colleagues! My name is Kireev Maxim Vladimirovich. I am a professor at the Saint-Petersburg State University and a Doctor of Science in biology. In this part of the course, we will discuss how tomographic technologies are used in the research of cerebral mechanisms of language. We will use as an example the research of the mental lexicon. Speaking of tomographic technologies, it needs to be pointed out that, compared to electrophysiological methods such as electroencephalography, and evoked potentials tomographic technologies allow us to precisely locate the cerebral structures that control the activity we are studying. Compared to tomographic methods, encephalography, despite having a good temporal resolution, cannot tell us exactly which cerebral structures are active at this very moment. When studying brain and cerebral mechanisms, it is important to know, firstly, which cerebral structures are active, and secondly, if they interact with each other and how exactly they interact. This enables us to research cerebral mechanisms underlying the activity we are studying, in this case, speech. In this slide, you can see various pictures made using the tomographic method. It also needs to be pointed out that, among tomographic methods, the most widely used for researching cerebral functions are positron-emission tomography and functional magnetic resonance imaging. To learn which cerebral structures are active, they use the fact that the blood flow in the brain is dependent on neuronal activity. Various methods of positron-emission tomography use what is called detection of radiopharmaceutical distribution. PET shows us how biologically active chemicals marked by radioactive isotopes are distributed throughout the brain. This is done to study blood flow, which enables us to indirectly assess changes in neuronal activity. To do this, researchers use the radiopharmaceutical H2O15, water labelled by oxygen-15. The PET method is convenient because using certain radiopharmaceuticals you can trace any biochemical process in the brain. For example, one can trace the metabolism of glucose or observe how the amino acid metabolism is distributed. However, to study cerebral functions and to see which cerebral structures are active, we use the radiopharmaceutical H2O15, water labelled by oxygen. What is the principal basis for positron-emission tomography? The point is that radiopharmaceuticals contain unstable atoms that decay, all the while emitting positrons. These positrons enter the body, collide with electrons in its material and annihilate that is, the positron and the electron destroy each other. This annihilation is followed by the emission of two gamma-rays that escape at an angle of exactly 180 degrees to each other. It is these instances of radioactive decay that are registered by the detectors of a PET scanner, that looks like a ring of detectors. It is important that an instance of such radioactive decay is only registered if two gamma-rays simultaneously reach the detectors that are located directly opposite to each other. In modern PET methods this is done using electronics, and it is called "autoelectronic collimation". It is a special mode that PET scanners operate in that allows to register gamma-rays in a relatively short time interval. Thus, the more rings a detector has, and the more compact they are, the better is its spatial resolution. This way, we get a pseudo-3D image that shows how the labelled chemical is distributed. Keeping in mind that we are talking about radiopharmaceuticals and, especially , those with a short half-life, there are time limits inherent in the design of such PET research. In this slide, you can see an example of how it looks. Let us say that there are two experimental conditions the first is control, and the second is a language-based task, in which the subjects are asked to name objects they see in the pictures. One scan that registers the redistribution of blood flow takes up from around 60 to 90 seconds. The pauses between such scans take up 30 to 40 minutes. The reason for this is that the subsequent scan can only be started after the complete decay of the dose of the radiopharmaceutical that the subject was injected with, which is usually about 1 milliliter. The half-life of water labelled by oxygen-15 is usually about 2 minutes Thus, one such experiment can last as long as several hours. Nowadays, positron-emission tomography techniques are losing out to magnetic resonance imaging that, firstly, doesn’t require such long breaks between scans, and, secondly, doesn’t expose subjects to any radiation. Nevertheless, PET can still be used in subjects with contraindications to functional MRI, for example, metal implants inside or on the body, or with medical contraindications to being exposed to powerful magnetic fields. Functional magnetic resonance imaging registers the so-called “BOLD signal”. This signal depends on the oxygenation of the subject’s blood. This is because small metal particles, when placed in range of the fMRI scanner, distort the signal in such a way that at certain moments the measured magnetic resonance signal from the areas where such metal is located, is substantially lower. When a large enough group of neurons is activated in a certain area of the brain, the area consumes more oxygen, and there is an increase in hemoglobin molecules concentration. There are the molecules that have given away their oxygen to the neurons so that the neurons can function more effectively. In these beginning moments, for example, when subject has just been shown a stimulus or has just started performing a task, the proportion of oxidized and unoxidized hemoglobin causes the signal that is being registered to fade. However, by way of compensation. there is an increase in blood flow in the area where the neuronal activity has been initiated, so that the hemoglobin that has lost its oxygen is washed away, giving way to oxidized hemoglobin, that is, hemoglobin carrying oxygen. And this, inversely, leads to an increase in the signal’s strength. This way, by registering the BOLD signal, we can indirectly demonstrate a local increase in neuronal activation in this area of the brain. As we know exactly what is happening and when during the scanning, this technique enables us to tell exactly which the areas of the brain start working or are activated at a given moment. The BOLD signal is known to have certain properties. You can see how it looks in this slide. At the start of neuronal activation, there is a slight decrease in the BOLD signal, then the signal increases again before levelling off. Usually, it peaks at around 6 seconds These properties of the BOLD signal are used in statistical analysis of individual MRI data, which helps us to model how the signal in certain areas will look when the subject is presented with the stimuli that are processed by those areas. This way, using fairly complex analytical methods, we can get statistical parameters showing how accurately the proposed statistical model predicts the changes in the BOLD signal. It is these statistical parameters that are later used in the analysis of group data. The methods of analysis of group data in both PET and MRI studies have certain similarities that we are going to talk about later. In this slide, you can see an example of the so-called “block design” of MRI studies, when the control and experimental stimuli are grouped in something like blocks. You can see that, in contrast to PET studies, we are able to present the subjects with stimuli without interruption. In this way, functional MRI methods have better time resolution and, as a result, they provide us with more information than PET methods. However, there are designs in which the experimental stimuli are presented more sparsely or in a random order. This is called “event-related design”, and it helps us locate the cerebral structures that process certain stimuli. Group data is analyzed in the same way for every experimental design, using the so-called “logic of activation research”. You can see in the slide an example of such an experiment, in which the experimental task is to name the pictures, and the control stimuli a simple checkerboard. The BOLD signal is accumulated for some time, and after that our task is to contrast, or compare, the experimental and control conditions. You can see in the slide that, in this kind of comparison, sensory data is automatically excluded and what is left are the brain areas associated with the process that we are studying, in this case - naming of objects in pictures. So, the clusters of voxels that are obtained in this way reflect precisely those cerebral structures that are active at the moment. It should also be noted that such comparison is done voxel-by-voxel, that is, each voxel is statistically analysed to learn if there are differences between two experimental conditions. This gives rise to what is called the “multiple comparisons problem”. When we repeatedly apply the same test to the same data, the probability of a false-positive result rises dramatically just because we use the 5% level of statistical significance. A good illustration of what may result from this property if we use such statistical methods is shown in this slide. This is a famous example in which the so-called “uncorrected image” during an fMRI study showed activation in a salmon’s brain when it was shown emotional pictures from human life. The important feature of this study was that the salmon was dead. This somewhat grotesque example demonstrates how important it is to use special correction methods for multiple comparisons to be able to successfully use such statistical methods.