The renaissance of functional 18F-FDG PET brain activation imaging

For a few years now, positron emission tomography (PET) has benefited from a renaissance of functional imaging through F-FDG PET brain activation studies. The first PET studies involving activation brain imaging nevertheless emerged in the 1980s, using intravenously administered oxygen-15labelled water (O-water) [1, 2]. This approach was used to measure regional cerebral blood flow (rCBF), and has been shown to be a sensitive method for quantifying regional brain activation during specific tasks [3, 4]. Such functional imaging studies proved extremely useful for mapping brain activation patterns involved in cognitive tasks such as word reading, mental imagery, timing or memory [5, 6]. The first clinical validations of quantitative analyses using Statistical Parametric Mapping (SPM, Wellcome Trust Centre for Neuroimaging, London, United Kingdom) were thus performed with PET imaging [7]. These PET imaging studies preceded the emergence of functional magnetic resonance imaging (fMRI) in the 1990s, which depicts the engagement of different brain regions within a distributed system through fluctuations of the blood-oxygen-level dependent (BOLD) signal [8]. fMRI applications have since been extended to reveal additional information about the degree to which components of large-scale neural systems are functionally coupled together to achieve specific tasks. This phenomenon underpins the study of functional connectivity, which is mathematically defined as the statistical association between two distinct time-series, i.e. the connectivity between brain regions that share functional properties. The fMRI approach enables the study of functional connectivity within single subjects or groups through serial imaging of the brain in the socalled resting state (i.e. without any specific stimulus or task) or in a condition of task-dependent activation [9]. In the past two decades MRI has largely supplanted PET for studies of brain activation, due to the lack of exposure to ionizing radiation and because of the superior spatial and especially temporal resolution, which allows for dynamic measurement of multiple responses within a single imaging session [8]. As such, fMRI has substantially advanced our understanding of normal and pathological brain functions, whether in resting-state or activation functional imaging studies [10, 11]. Nevertheless, fMRI does have some disadvantages. At the single-subject level, fMRI exhibits low signal-to-noise ratios and low variance concentrations in comparison with PET [12]. Indeed, to represent a given aggregate percentage of variance, many more components are needed for fMRI than for PET. Therefore, multiple runs of the same fMRI protocol are required to increase the sensitivity of measurements. Furthermore, recent work has shown that the signal-to-noise ratio may vary dramatically between fMRI runs, especially when performing a brief task, which degrades the robustness and reproducibility of this method at a single-subject level [13]. All these disadvantages generate poor out-of-sample replications in fMRI studies. In addition, fMRI is vulnerable to ferromagnetic artefacts, which are especially prominent in anterior and ventral brain areas [14]. These limitations could also cause others difficulties in performing such imaging due to the high prevalence of treatment with implantable devices for neurostimulation, most of which are not MR compatible [15]. Noise generated by the gradient coils, which could adversely affect listening to auditory inputs and can make recording of spoken outputs difficult, or movement artefacts that arise from speaking or other orofacial movements, could also cause others technical limitations [14]. Furthermore, the BOLD signal is a composite of effects attributable to the * Eric Guedj eric.guedj@ap-hm.fr