NeuroImage
-
Machine learning is increasingly being applied to neuroimaging data. However, most machine learning algorithms have not been designed to accommodate neuroimaging data, which typically has many more data points than subjects, in addition to multicollinearity and low signal-to-noise. Consequently, the relative efficacy of different machine learning regression algorithms for different types of neuroimaging data are not known. ⋯ Random Forest also produced a moderate performance for small effect sizes, but could do so across all sample sizes. Machine learning techniques also improved prediction accuracy for multiple regression. These data provide empirical evidence for the differential performance of various machines on neuroimaging data, which are dependent on number of sample size, features and effect size.
-
Brain arteriovenous malformations (AVMs) are congenital vascular anomalies characterized by arteriovenous shunting through a network of coiled and tortuous vessels. Because of this anatomy, the venous drainage of an AVM is hypothesized to contain more oxygenated, arterialized blood than healthy veins. By exploiting the paramagnetic properties of deoxygenated hemoglobin in venous blood using magnetic resonance imaging (MRI) quantitative susceptibility mapping (QSM), we aimed to explore venous density and oxygen saturation (SvO2) in patients with a brain AVM. ⋯ Therefore, QSM can be used to detect SvO2 alterations in brain AVMs. However, since factors such as flow-induced signal dephasing or the presence of hemosiderin deposits also strongly affect QSM image contrast, AVM vein segmentation must be performed based on alternative MRI acquisitions, e.g., time of flight magnetic resonance angiography or T1-weighted images. This is the first study to show, non-invasively, that AVM draining veins have a significantly larger SvO2 than healthy veins, which is a finding congruent with arteriovenous shunting.
-
With myelin playing a vital role in normal brain integrity and function and thus in various neurological disorders, myelin sensitive magnetic resonance imaging (MRI) techniques are of great importance. In particular, multi-exponential T2 relaxation was shown to be highly sensitive to myelin. The myelin water imaging (MWI) technique allows to separate the T2 decay into short components, specific to myelin water, and long components reflecting the intra- and extracellular water. ⋯ In MRI, iron extraction lead to a decrease in MWF by 26%-28% in WM. Thus, a change in MWF does not necessarily reflect a change in myelin content. This observation has important implications for the interpretation of MWI findings in previously published studies and future research.
-
Quantitative Susceptibility Mapping (QSM) provides a way of measuring iron concentration and myelination non-invasively and has the potential of becoming a tool of paramount importance in the study of a host of different pathologies. However, several experimental factors and the physical properties of magnetic susceptibility (χ) can impair the reliability of QSM, and it is therefore essential to assess QSM reproducibility for repeated acquisitions and different field strength. ⋯ To maximize intra-scanner reproducibility it is necessary to match the TEs of the acquisition protocol, but the application of this rule leads to inconsistent QSM values across scanners operating at different static magnetic field. This study however demonstrates that, provided a careful choice of acquisition parameters, and in particular by using TEs at 3T that are approximately 2.6 times longer than those at 7T, highly reproducible whole-brain χ maps can be achieved also across different scanners, which renders QSM a suitable technique for longitudinal follow-up in clinical settings and in multi-center studies.
-
In the present study, we report the results from a large sample of participants (N = 136), selected based on their EEG quality, to obtain event-related potential (ERP) normative data. All participants were tested in Simple Response Task (SRT) and Discriminative Response Task (DRT). A subset of 36 participants was tested also in Passive Vision task. ⋯ Spatiotemporal patterns of all the observed components were analyzed using source analysis. Beside the well-known ERP components, we also described recently identified prefrontal components: the pre-stimulus prefrontal negativity (pN) associated to proactive cognitive (mainly inhibitory) control within the inferior frontal gyrus (iFg); the post-stimulus prefrontal N1, P1 and P2 (pN1, pP1 and pP2) involved in perceptual and visual-motor awareness (pN1 and pP1), and in stimulus-response mapping and decision-making (pP2) localized within the insular cortex. The large sample of high-quality EEG datasets allowed to identify four additional components: the pre-stimulus visual negativity (vN) originating in extrastriate visual areas and interpreted as a visual readiness activity; the post-stimulus prefrontal N2 and N3 (pN2 and pN3) components interpreted as feedback reactivation of the anterior insular cortex; and the post-stimulus prefrontal P3 (pP3), interpreted as persisting inhibitory activity of the iFg for inhibited trials.