Functional and Molecular Neuroimaging

Functional Magnetic Resonance Imaging

Today, fMRI is a standard technique in neuroscience brain imaging. It relates to the blood oxygen level–dependent (BOLD) effect, which is due to a transient and local excess of oxygenated blood resulting from changes in regional CBF and neuronal activity. Oxygenated hemoglobin is used here as an intrinsic contrast agent and serves as a surrogate marker of neuronal oxygen consumption and activity.

One approach to studying the integrity of functional neuronal networks uses experimental stimuli (e.g., words that have to be read) either in a block design (series of words for 20–30 seconds alternating by rest blocks of similar length over several minutes) or event-related tests (≈30–40 stimuli of each type presented in a counterbalanced order, each followed by some baseline period). Experiments are often conducted with multiple subjects, which requires stereotactic normalization into a standard space. Time series are analyzed using univariate analyses within the general linear model (GLM), enabling inferences on local effect sizes. Resulting visualizations illustrate regions with task-specific statistically significant differences in brain activation. More recently multivariate analyses, such as partial least squares, enable inferences on network connectivity on the whole-brain level. Complementary approaches use graph theory analysis to elucidate the network features during task condition or resting state (see later) and their alteration through brain disease or assessment of effective connectivity using causal inference modes, such as dynamic causal modeling, Granger causality, or Bayesian learning networks.

Parallel to task-related MRI, resting-state fMRI (Rs-fMRI) has been developed for applications in dementia research ( ). Analysis of spontaneous fluctuations of the BOLD signal during resting-state conditions has revealed consistent networks of intrinsic connectivity that partly map with functional networks activated during task performance, such as motor networks, language networks, attention networks, or deactivated during task performance, such as the default mode network (DMN) involving medial temporal lobe, superior parietal, and prefrontal lobe areas. Analysis of resting-state intrinsic connectivity networks typically involves analysis of correlations with seed regions or network analysis using multivariate techniques such as independent component analysis. Rs-fMRI analyses have been used in the analysis of prodromal or manifest stages of dementia due to the more limited requirements for patients’ compliance in comparison with task-based fMRI. However, the high interscanner and longitudinal variability of resting-state networks (in comparison with task-elicited functional networks) limits their utility for individual diagnostic or prognostic applications.

Arterial Spin Labeling

Arterial spin labeling (ASL) is an MRI technique that provides estimates of cerebral perfusion at the tissue level noninvasively and without the administration of contrast media. The main physiological parameter that can be measured by ASL is the CBF. ASL imaging techniques provide quantitative parametric imaging maps of CBF for visual and region-of-interest (ROI)–based analysis ( ).

ASL was first introduced in the early 1990s ( ). However, its main drawback is its inherent low signal-to-noise ratio (SNR; ). Although ASL is possible with 1,5 tesla (T) MR systems, low SNR increases the necessary scan time and therefore makes the technique sensitive to motion artifacts. Recent advances in coil technology and increasing field strength of the MR systems have led to a rapidly growing interest in ASL in clinical and preclinical imaging ( ).

In contrast to other MR-based techniques for the evaluation of tissue perfusion (e.g., dynamic contrast-enhanced MRI [DCE-MRI]), ASL uses the water molecules of the blood as an endogenous contrast agent to estimate tissue perfusion. ASL is based on the strategy of magnetically labeling the protons in blood molecules before they flow into the tissue of interest. According to the spin labeling technique used, ASL can be divided into three different types: pulsed ASL (PASL), pseudo-continuous ASL (pCASL), and continuous ASL (CASL). Currently the most used types of ASL are pCASL and PASL.

Important technical parameters for ASL acquisition are positioning of the labeling plane below the brain, labeling duration, and the postlabeling delay (PLD) or inflow-time of the postlabeling period. This delay describes the time between the end of the labeling period and the start of the imaging period. It describes the time allowed for the labeled blood to enter the tissue of interest within the imaging volume. The PLD depends on the blood velocity, which is correlated with the subject’s age. Because older patients have a decreased velocity, the recommended PLD is 1500 ms for pediatric patients and 1800 and 2000 ms for healthy adults below and above 70 years of age, respectively. For adult patients, a PLD of 2000 ms is recommended.

There is also a need for background suppression and prevention of patient motion during image acquisition to reduce noise and artifacts masking the signal difference, thus subsequently hindering image analysis. Image readout of ASL was traditionally based on fast echo planar imaging (EPI) techniques. Recently more advanced three-dimensional (3D) acquisition techniques have been proposed (e.g., 3D gradient and spin echo [GRASE] or 3D rapid acquisition relaxation enhanced [RARE] techniques). Compared with two-dimensional (2D) techniques, 3D readout has superior SNR and allows the acquisition of the entire volume of interest within one shot, thus reducing the slice-dependent variations of the perfusion signal observed in 2D techniques ( ).

Positron Emission Tomography

The concept of modern PET was developed during the 1970s ( ). The underlying principle of PET and also of SPECT is to image and quantify a physiological function or molecular target of interest in vivo by noninvasively assessing the spatial and temporal distribution of the radiation emitted by an intravenously injected or inhaled target-specific probe (radiotracer). Importantly, PET and SPECT tracers are administered in a nonpharmacological dose (micrograms or less), so they neither perturb the underlying system nor cause pharmacological effects. Because of their ability to enable the visualization of molecular targets and functions on a macroscopic level with unsurpassed sensitivity down to a picomolar concentration, PET and SPECT are also called molecular imaging techniques. (See , for a textbook on PET and SPECT physics.) (See Table 42.1 for a glossary of PET and SPECT tracers.)

In the case of PET, a positron-emitting radiopharmaceutical is injected or inhaled by the subject. The emitted positron travels a short distance in tissue (effective range <1 mm for common PET radionuclides) before it encounters an electron; the positron and electron combine, yielding a pair of two 511-keV annihilation photons emitted in opposite directions and detected quasi-simultaneously by a PET detector ring surrounding the patient. 3D PET image datasets of the distribution of the PET tracer and its target are then generated by standard image reconstruction algorithms. To actually gain quantitative PET images (i.e., radioactivity/tracer concentration per unit tissue), the acquired data are corrected for scatter and random coincidences and photon attenuation by tissue absorption (e.g., using low-dose CT in the case of a PET/CT scanner). The spatial resolution of modern PET systems is about 3–5 mm. Thus PET is susceptible to partial volume effects if the size of the object or lesion is less than twice the scanner’s resolution (as a rule of thumb). Today’s PET systems are either constructed as hybrid PET/CT or, more recently, PET/MRI systems. Although the clinical utility of the latter still needs to be defined, integrated PET/MRI allows for the comprehensive, synchronous imaging of multiple morphological, functional, and molecular parameters in a single scanning session.

Commonly used radionuclides in neurological PET research studies are carbon-11 ( ¹¹ C, half-life = 20.4 minutes), oxygen-15 ( ¹⁵ O, physical half-life = 2.03 minutes), and fluorine-18 ( ¹⁸ F, half-life = 109.7 minutes), which are all produced in cyclotrons. Although the relatively long half-life of ¹⁸ F allows for the shipping of ¹⁸ F-labeled tracers from a cyclotron site to a distant clinical PET site, this is not possible for shorter-lived isotopes. Thus only ¹⁸ F-labeled radiopharmaceuticals are widely available and have been used clinically.

The most frequently used PET radiopharmaceutical is the glucose analog [ ¹⁸ F]2-deoxy-2-fluoro-d-glucose ([ ¹⁸ F]FDG). It was originally developed to assess cerebral glucose metabolism and is now also widely used as a tumor imaging agent in oncology. With the rate of glucose metabolism being closely related to maintenance of ion gradients and transmitter turnover (in particular, glutamate), [ ¹⁸ F]FDG represents an ideal tracer for the assessment of neuronal function and its changes ( ). After uptake in cerebral tissue by specific glucose transporters, [ ¹⁸ F]FDG is phosphorylated by hexokinase. Since [ ¹⁸ F]FDG-6-P is not a substrate for transport out of the cell and cannot be metabolized further, it is irreversibly trapped in cells. Therefore the distribution of [ ¹⁸ F]FDG in tissue imaged by PET (started 30–60 minutes after injection to allow for sufficient uptake; 5–20 minute scan duration) closely reflects the regional distribution of cerebral glucose metabolism. It is highest in neocortical areas, basal ganglia, and thalamus; intermediate in hippocampus, amygdala, brainstem and cerebellum; and lowest in white matter. There is an age-related decline of cerebral glucose metabolism, most prominently in frontal cortex ( ). For clinical use, normal plasma glucose levels and standard resting-state conditions (low noise level and low ambient light) should be maintained during tracer accumulation to avoid changes in the metabolic pattern by regional brain activation ( ).

Neuronal function is closely related to synaptic function, which involves the release of synaptic vesicles. Recently tracers have been developed to image the synaptic vesicle protein SV2A ( ). This could provide a molecular marker for synaptic function that is complementary to functional imaging with [ ¹⁸ F]FDG and fMRI.

A large number of other PET tracers have been used in clinical research over the past decades, and a few have established clinical diagnostic utility. They are mentioned in the respective disease chapters. Generally the clinical interpretation of PET images is done by qualitative visual inspection, often supported by voxel-wise statistical analyses in comparison to normal subjects. However, PET also provides quantitative regional data, which are mainly used for research. The highest levels of quantitative accuracy employ mathematical modeling of tracer kinetics in tissue in relation to an arterial or reference tissue input function. However, such research techniques require dedicated expertise, software, and often also equipment for measuring plasma activity and tracer metabolites ( ) and are not usually required for the clinical use of PET.

Single-Photon Emission Computed Tomography

The first SPECT measurements were performed in the 1960s ( ). SPECT employs gamma-emitting radionuclides that decay by emitting a single gamma ray. Typical radionuclides employed for neurological SPECT are technetium-99m ( ^99m Tc; half-life = 6.02 hours) and iodine-123 ( ¹²³ I; half-life = 13.2 hours). Gamma cameras are used for SPECT acquisition, whereby usually two detector heads rotate around the patient’s head to acquire 2D planar images (projections) of the head from multiple angles (e.g., in three-degree steps). Whereas radiation collimation is achieved by coincidence detection in PET, hardware collimators with lead septa are placed in front of the detector heads in the case of SPECT scanners. Finally, 3D image data reconstruction is done by conventional reconstruction algorithms. With combined SPECT/CT systems, a low-dose CT transmission scan can replace less accurate calculated attenuation correction.

The different acquisition principles imply that SPECT possesses a considerably lower sensitivity than PET. Thus rapid temporal sampling (image frames of seconds to minutes) as a prerequisite for pharmacokinetic analyses is the strength of PET, whereas a single SPECT acquisition usually takes 20–30 minutes. Furthermore, the spatial resolution of modern SPECT is only about 7–10 mm, deteriorating with increasing distance between object and collimator (i.e., higher resolution for cortical than subcortical structures; distance between patient and collimator should be minimized for optimal resolution). Thus SPECT is more susceptible to partial volume effects than PET, which can be a particular drawback when it comes to imaging small structures or lesions (e.g., brain tumors). Nevertheless, brain-dedicated SPECT instruments have been proposed that allow for optimized spatial and temporal sampling and pharmacokinetic data quantification ( ). The important advantages of SPECT over PET are the lower costs and broader availability of SPECT systems and radionuclides. Although ¹²³ I-labeled tracers (e.g., [ ¹²³ I]FP-CIT ([ ¹²³ I]ioflupane, DaTSCAN) for dopamine transporter [DAT] imaging) can easily be shipped over long distances, technetium-99m can be eluted onsite from molybdenum-99 ( ⁹⁹ Mo)/ ^99m Tc generators and used for labeling commercially available radiopharmaceutical kits.

The two most widely used CBF tracers are hexamethylpropyleneamine oxime ([ ^99m Tc]HMPAO) and ethylcysteinate dimer ([ ^99m Tc]ECD). Owing to their lipophilic nature and thus high first-pass brain extraction, both radiotracers are rapidly taken up by the brain. They are quasi-irreversibly retained after conversion into hydrophilic compounds (enzymatic deesterification of [ ^99m Tc]ECD; instability, and possibly interaction with glutathione in the case of [ ^99m Tc]HMPAO). Differences in uptake mechanisms may explain slight differences in biological behavior (e.g., in stroke), with [ ^99m Tc]HMPAO being more closely correlated to perfusion, whereas [ ^99m Tc]ECD uptake is also influenced by metabolic activity. Despite the fact that cerebral radiotracer uptake is virtually complete within just 1–2 minutes of injection, SPECT acquisition is usually started after 30–60 minutes to allow for sufficient background clearance.

Given the fact that the CBF is closely coupled to cerebral glucose metabolism and thus to neuronal function (with a few rare exceptions), [ ^99m Tc]HMPAO and [ ^99m Tc]ECD are used to assess neuronal activity indirectly. However, since cerebral autoregulation is also affected by many other factors (e.g., carbon dioxide level) and possibly diseases, cerebral glucose metabolism represents a more direct and less variable marker of neuronal activity. Given the technical limitations mentioned earlier, [ ¹⁸ F]FDG PET is generally preferred to CBF SPECT; therefore the following sections of this chapter primarily refer to PET studies. One important exception, however, is the use of ictal CBF SPECT in the assessment of patients with epilepsy.

Overview of Dementia

Dementia is one of the main challenges to human health. Its importance is increasing worldwide due to the overall increase in life expectancy. Although the diagnosis of dementia is made by clinical assessment, the diagnosis of specific diseases causing dementia increasingly relies on imaging. The most prevalent cause of dementia is Alzheimer disease, which is characterized by deposits of the pathological proteins beta-amyloid and tau. Current research criteria for AD define the presence of the disease based on the detection of specific biomarkers (i.e., amyloid accumulation and tau) as detected in CSF or by PET ( ). Notably, under this definition, Alzheimer disease (AD) does not imply presence of dementia at the time of diagnosis but indicates a disease process that will eventually lead to Alzheimer dementia. A positive test for amyloid and tau in the absence of cognitive impairment is an indicator of preclinical AD, or of prodromal AD, known as mild cognitive impairment (MCI). These conceptual developments can rely on molecular imaging techniques based on PET disease-defining biomarkers of AD. Thus detection of amyloid deposits in vivo by PET enables earlier diagnosis and improved distinction from other diseases causing dementia, such as frontotemporal dementia (FTD), dementia with Lewy bodies (DLB), and vascular dementia (VD). The different patterns of cortical functional deficits associated with these diseases are detected by functional imaging techniques, with [ ¹⁸ F]FDG PET being the most firmly established technique and various methods to assess CBF (SPECT, PET, ASL MRI) as potential alternatives. Typically these deficits also affect communication between brain areas, which increasingly is being studied by techniques for the analysis for brain networks, in particular by resting-state fMRI. PET and SPECT imaging of specific transmitter systems, which are impaired in AD (cholinergic) and DLB (dopaminergic and cholinergic), can provide further information for diagnosis, prediction, and the monitoring of therapy.

Alzheimer Disease

[ ¹⁸ F]2-deoxy-2-fluoro-d-glucose Positron Emission Tomography

The early and accurate diagnosis of dementia is of crucial importance for appropriate treatment (including possible enrollment into clinical trails and avoidance of possible side effects of treatments), for prognosis, and for adequate counseling of patients and caregivers. The diagnostic power of [ ¹⁸ F]FDG PET in this situation is well established ( ) and reflected in clinical guidelines ( ). [ ¹⁸ F]FDG PET has also been included as a biomarker of neuronal damage in current diagnostic research criteria ( ). In clinical practice, [ ¹⁸ F]FDG PET studies are interpreted by qualitative visual readings. To achieve optimal diagnostic accuracy, these readings should be supported by voxel-based statistical analyses in comparison with aged-matched normal controls ( ). PET studies should always be interpreted with parallel inspection of a recent CT or MRI scan to detect structural defects (e.g., ischemia, atrophy, subdural hematoma), which can cause regional hypometabolism.

The typical finding in AD dementia is bilateral hypometabolism of the temporal and parietal association cortices, with the temporoparietal junction being the center of impairment. As the disease progresses, frontal association cortices also become involved ( Figs. 42.1 and 42.2 ). The magnitude and extent of the hypometabolism increase with progressing disease, although there is relative sparing of the primary motor and visual cortices, basal ganglia, and cerebellum (often used as reference regions). The degree of hypometabolism is closely correlated with the severity of dementia ( ). Cortical hypometabolism is bilateral but often asymmetrical, corresponding to predominant clinical symptoms (language impairment if the dominant hemisphere is most affected or visuospatial impairment if the nondominant hemisphere is most affected). Voxel-based statistical analyses consistently show that the posterior cingulate gyrus and precuneus are also affected; this is an important diagnostic clue even in the early stages of AD ( ). The hippocampus is most affected by AD pathology, in particular by pathological tau deposits and atrophy, but hypometabolism appears less pronounced than in temporoparietal cortex ( ). This may be due to limitations of PET spatial resolution, although relative functional hyperactivity has also been discussed ( ). A clear reduction of hippocampal metabolism in patients was observed as compared with selected control patients who remained stable over at least 4 subsequent years ( ).

The logopenic variant, primary progressive aphasia (lvPPA), which is characterized by most prominent deficits in word retrieval and sentence repetition, is commonly also caused by AD ( ). LvPPA patients typically show a strongly leftward asymmetric hypometabolism of the temporoparietal cortex ( ; Fig. 42.3 ). Patients with posterior cortical atrophy (PCA), another nonamnestic presentation of AD with predominant visuospatial and visuoperceptual deficits, typically exhibit pronounced hypometabolism of occipital association cortex ( ; Fig. 42.4 ).

A meta-analysis of [ ¹⁸ F]FDG-PET cross-sectional case-control studies ( n = 562 in total) revealed a very high sensitivity (96%) and specificity (90%) of [ ¹⁸ F]FDG PET for the diagnosis of AD in patients with dementia ( ). In [ ¹⁸ F]FDG PET studies with autopsy confirmation in patients with memory complaints, the pattern of temporoparietal hypometabolism as assessed by visual readings alone showed a high sensitivity of 84%–94% for detecting pathologically confirmed AD, with a specificity of 73%–74% ( ). Visual inspection of [ ¹⁸ F]FDG PET was found to be of similar accuracy as a clinical follow-up examination performed 4 years after PET. Moreover, when [ ¹⁸ F]FDG PET disagreed with the initial clinical diagnosis, the PET diagnosis was considerably more likely to be congruent with the pathological diagnosis than the clinical diagnosis ( ).

Patterns of hypoperfusion observed with CBF SPECT in AD are similar to metabolic deficits on [ ¹⁸ F]FDG PET; but according to meta-analyses ( ) and direct comparisons ( ), their accuracy is inferior. There are increasing numbers of techniques to provide CBF mapping as replacement of FDG PET, including ASL MR and imaging early uptake of amyloid and tau tracers (see the following section). Studies in selected patient series suggest similar accuracy as [ ¹⁸ F]FDG PET, but larger studies are still needed to confirm the robustness of these findings.