Machine learning and artificial intelligence in neuroscience: A primer for researchers.

Artificial intelligence (AI) is often used to describe the automation of complex tasks that we would attribute intelligence to. Machine learning (ML) is commonly understood as a set of methods used to develop an AI. Both have seen a recent boom in usage, both in scientific and commercial fields. For the scientific community, ML can solve bottle necks created by complex, multi-dimensional data generated, for example, by functional brain imaging or *omics approaches. ML can here identify patterns that could not have been found using traditional statistic approaches. However, ML comes with serious limitations that need to be kept in mind: their tendency to optimise solutions for the input data means it is of crucial importance to externally validate any findings before considering them more than a hypothesis. Their black-box nature implies that their decisions usually cannot be understood, which renders their use in medical decision making problematic and can lead to ethical issues. Here, we present an introduction for the curious to the field of ML/AI. We explain the principles as commonly used methods as well as recent methodological advancements before we discuss risks and what we see as future directions of the field. Finally, we show practical examples of neuroscience to illustrate the use and limitations of ML.

Study protocol: effects of treatment expectation toward repetitive transcranial magnetic stimulation (rTMS) in major depressive disorder-a randomized controlled clinical trial.

BACKGROUND: Patients' expectations toward any given treatment are highly important for the effectiveness of such treatment, as has been demonstrated for several disorders. In particular, in major depressive disorder (MDD), one of the most frequent and most serious mental disorders with severe consequences for the affected, the augmentation of available treatment options could mean a ground-breaking success. Repetitive transcranial magnetic stimulation (rTMS), a new, non-invasive, and well-tolerated intervention with proven effects in the treatment of MDD, appears particularly suitable in this context as it is assumed to exert its effect via structures implicated in networks relevant for both expectation and depression. METHODS: All patients will receive rTMS according to its approval. Half of the patients will be randomized to a psychological intervention, which is a comprehensive medical consultation aiming to improve positive treatment expectations; the control group will receive a conventional informed consent discussion (in the sense of a treatment-as-usual condition). As outcome parameters, instruments for both self-assessment and external assessment of depression symptoms will be applied. Furthermore, psycho-immunological parameters such as inflammation markers and the cortisol awakening response in saliva will be investigated. Resting-state functional magnetic resonance imaging (rs fMRI) will be performed to analyze functional connectivity, including the cerebellum, and to identify neuronal predictors of expectation effects. In addition, possible cerebellar involvement will be assessed based on a cerebellar-dependent motor learning paradigm (i.e., eyeblink conditioning). DISCUSSION: In this study, the effects of treatment expectations towards rTMS are investigated in patients with MDD. The aim of this study is to identify the mechanisms underlying the expectation effects and, beyond that, to expand the potential of non-invasive and well-tolerated treatments of MDD. TRIAL REGISTRATION: German Registry of Clinical Studies (DRKS DRKS00028017. Registered on 2022/03/07. URL: https://www.drks.de/drks_web/ .

Brain morphology predicts individual sensitivity to pain: a multicenter machine learning approach.

ABSTRACT: Sensitivity to pain shows a remarkable interindividual variance that has been reported to both forecast and accompany various clinical pain conditions. Although pain thresholds have been reported to be associated to brain morphology, it is still unclear how well these findings replicate in independent data and whether they are powerful enough to provide reliable pain sensitivity predictions on the individual level. In this study, we constructed a predictive model of pain sensitivity (as measured with pain thresholds) using structural magnetic resonance imaging-based cortical thickness data from a multicentre data set (3 centres and 131 healthy participants). Cross-validated estimates revealed a statistically significant and clinically relevant predictive performance (Pearson r = 0.36, P < 0.0002, R2 = 0.13). The predictions were found to be specific to physical pain thresholds and not biased towards potential confounding effects (eg, anxiety, stress, depression, centre effects, and pain self-evaluation). Analysis of model coefficients suggests that the most robust cortical thickness predictors of pain sensitivity are the right rostral anterior cingulate gyrus, left parahippocampal gyrus, and left temporal pole. Cortical thickness in these regions was negatively correlated to pain sensitivity. Our results can be considered as a proof-of-concept for the capacity of brain morphology to predict pain sensitivity, paving the way towards future multimodal brain-based biomarkers of pain.

Predictive modeling of optimism bias using gray matter cortical thickness.

People have been shown to be optimistically biased when their future outcome expectancies are assessed. In fact, we display optimism bias (OB) toward our own success when compared to a rival individual's (personal OB [POB]). Similarly, success expectancies for social groups we like reliably exceed those we mention for a rival group (social OB [SOB]). Recent findings suggest the existence of neural underpinnings for OB. Mostly using structural/functional MRI, these findings rely on voxel-based mass-univariate analyses. While these results remain associative in nature, an open question abides whether MRI information can accurately predict OB. In this study, we hence used predictive modelling to forecast the two OBs. The biases were quantified using a validated soccer paradigm, where personal (self versus rival) and social (in-group versus out-group) forms of OB were extracted at the participant level. Later, using gray matter cortical thickness, we predicted POB and SOB via machine-learning. Our model explained 17% variance (R2 = 0.17) in individual variability for POB (but not SOB). Key predictors involved the rostral-caudal anterior cingulate cortex, pars orbitalis and entorhinal cortex-areas that have been associated with OB before. We need such predictive models on a larger scale, to help us better understand positive psychology and individual well-being.

Statistical quantification of confounding bias in machine learning models.

BACKGROUND: The lack of nonparametric statistical tests for confounding bias significantly hampers the development of robust, valid, and generalizable predictive models in many fields of research. Here I propose the partial confounder test, which, for a given confounder variable, probes the null hypotheses of the model being unconfounded. RESULTS: The test provides a strict control for type I errors and high statistical power, even for nonnormally and nonlinearly dependent predictions, often seen in machine learning. Applying the proposed test on models trained on large-scale functional brain connectivity data (N= 1,865) (i) reveals previously unreported confounders and (ii) shows that state-of-the-art confound mitigation approaches may fail preventing confounder bias in several cases. CONCLUSIONS: The proposed test (implemented in the package mlconfound; https://mlconfound.readthedocs.io) can aid the assessment and improvement of the generalizability and validity of predictive models and, thereby, fosters the development of clinically useful machine learning biomarkers.

Temporal dynamics of fMRI signal changes during conditioned interoceptive pain-related fear and safety acquisition and extinction.

Associative learning and memory mechanisms drive interoceptive signaling along the gut-brain axis, thus shaping affective-emotional reactions and behavior. Specifically, learning to predict potentially harmful, visceral pain is assumed to succeed within very few trials. However, the temporal dynamics of cerebellar and cerebral fMRI signal changes underlying early acquisition and extinction of learned fear signals and the concomitant evolvement of safety learning remain incompletely understood. 3 T fMRI data of healthy individuals from three studies were uniformly processed across the whole brain and the cerebellum. All studies employed differential delay conditioning (N = 94) with one visual cue (CS+) being repeatedly paired with visceral pain as unconditioned stimulus (US) while a second cue remained unpaired (CS-). During subsequent extinction (N = 51), all CS were presented without US. Behavioral results revealed increased CS+-aversiveness and CS--pleasantness after conditioning and diminished valence ratings for both CS following extinction. During early acquisition, the CS- induced linearly increasing neural activation in the insula, midcingulate cortex, hippocampus, precuneus as well as cerebral and cerebellar somatomotor regions. The comparison between acquisition and extinction phases yielded a CS--induced linear increase in the posterior cingulate cortex and precuneus during early acquisition, while there was no evidence for linear fMRI signal changes for the CS+ during acquisition and for both CS during extinction. Based on theoretical accounts of discrimination and temporal difference learning, these results suggest a gradual evolvement of learned safety cues that engage emotional arousal, memory, and cortical modulatory networks. As safety signals are presumably more difficult to learn and to discriminate from learned threat cues, the underlying temporal dynamics may reflect enhanced salience and prediction processing as well as increasing demands for attentional resources and the integration of multisensory information. Maladaptive responses to learned safety signals are a clinically relevant phenotype in multiple conditions, including chronic visceral pain, and can be exceptionally resistant to modification or extinction. Through sustained hypervigilance, safety seeking constitutes one key component in pain and stress-related avoidance behavior, calling for future studies targeting the mechanisms of safety learning and extinction to advance current cognitive-behavioral treatment approaches.

Resting-state functional heterogeneity of the right insula contributes to pain sensitivity.

Previous studies have described the structure and function of the insular cortex in terms of spatially continuous gradients. Here we assess how spatial features of insular resting state functional organization correspond to individual pain sensitivity. From a previous multicenter study, we included 107 healthy participants, who underwent resting state functional MRI scans, T1-weighted scans and quantitative sensory testing on the left forearm. Thermal and mechanical pain thresholds were determined. Connectopic mapping, a technique using non-linear representations of functional organization was employed to describe functional connectivity gradients in both insulae. Partial coefficients of determination were calculated between trend surface model parameters summarizing spatial features of gradients, modal and modality-independent pain sensitivity. The dominant connectopy captured the previously reported posteroanterior shift in connectivity profiles. Spatial features of dominant connectopies in the right insula explained significant amounts of variance in thermal (R2 = 0.076; p < 0.001 and R2 = 0.031; p < 0.029) and composite pain sensitivity (R2 = 0.072; p < 0.002). The left insular gradient was not significantly associated with pain thresholds. Our results highlight the functional relevance of gradient-like insular organization in pain processing. Considering individual variations in insular connectopy might contribute to understanding neural mechanisms behind pain and improve objective brain-based characterization of individual pain sensitivity.

Network properties and regional brain morphology of the insular cortex correlate with individual pain thresholds.

Pain thresholds vary considerably across individuals and are influenced by a number of behavioral, genetic and neurobiological factors. However, the neurobiological underpinnings that account for individual differences remain to be fully elucidated. In this study, we used voxel-based morphometry (VBM) and graph theory, specifically the local clustering coefficient (CC) based on resting-state connectivity, to identify brain regions, where regional gray matter volume and network properties predicted individual pain thresholds. As a main finding, we identified a cluster in the left posterior insular cortex (IC) reaching into the left parietal operculum, including the secondary somatosensory cortex, where both regional gray matter volume and the local CC correlated with individual pain thresholds. We also performed a resting-state functional connectivity analysis using the left posterior IC as seed region, demonstrating that connectivity to the pre- as well as postcentral gyrus bilaterally; that is, to the motor and primary sensory cortices were correlated with individual pain thresholds. To our knowledge, this is the first study that applied VBM in combination with voxel-based graph theory in the context of pain thresholds. The co-location of the VBM and the local CC cluster provide first evidence that both structure and function map to the same brain region while being correlated with the same behavioral measure; that is, pain thresholds. The study highlights the importance of the posterior IC, not only for pain perception in general, but also for the determination of individual pain thresholds.

Deciphering the scopolamine challenge rat model by preclinical functional MRI.

During preclinical drug testing, the systemic administration of scopolamine (SCO), a cholinergic antagonist, is widely used. However, it suffers important limitations, like non-specific behavioural effects partly due to its peripheral side-effects. Therefore, neuroimaging measures would enhance its translational value. To this end, in Wistar rats, we measured whisker-stimulation induced functional MRI activation after SCO, peripherally acting butylscopolamine (BSCO), or saline administration in a cross-over design. Besides the commonly used gradient-echo echo-planar imaging (GE EPI), we also used an arterial spin labeling method in isoflurane anesthesia. With the GE EPI measurement, SCO decreased the evoked BOLD response in the barrel cortex (BC), while BSCO increased it in the anterior cingulate cortex. In a second experiment, we used GE EPI and spin-echo (SE) EPI sequences in a combined (isoflurane + i.p. dexmedetomidine) anesthesia to account for anesthesia-effects. Here, we also examined the effect of donepezil. In the combined anesthesia, with the GE EPI, SCO decreased the activation in the BC and the inferior colliculus (IC). BSCO reduced the response merely in the IC. Our results revealed that SCO attenuated the evoked BOLD activation in the BC as a probable central effect in both experiments. The likely peripheral vascular actions of SCO with the given fMRI sequences depended on the type of anesthesia or its dose.

Meta-analysis of neural systems underlying placebo analgesia from individual participant fMRI data.

The brain systems underlying placebo analgesia are insufficiently understood. Here we performed a systematic, participant-level meta-analysis of experimental functional neuroimaging studies of evoked pain under stimulus-intensity-matched placebo and control conditions, encompassing 603 healthy participants from 20 (out of 28 eligible) studies. We find that placebo vs. control treatments induce small, widespread reductions in pain-related activity, particularly in regions belonging to ventral attention (including mid-insula) and somatomotor networks (including posterior insula). Behavioral placebo analgesia correlates with reduced pain-related activity in these networks and the thalamus, habenula, mid-cingulate, and supplementary motor area. Placebo-associated activity increases occur mainly in frontoparietal regions, with high between-study heterogeneity. We conclude that placebo treatments affect pain-related activity in multiple brain areas, which may reflect changes in nociception and/or other affective and decision-making processes surrounding pain. Between-study heterogeneity suggests that placebo analgesia is a multi-faceted phenomenon involving multiple cerebral mechanisms that differ across studies.

Resection of cerebellar tumours causes widespread and functionally relevant white matter impairments.

Several diffusion tensor imaging studies reveal that white matter (WM) lesions are common in children suffering from benign cerebellar tumours who are treated with surgery only. The clinical implications of WM alterations that occur as a direct consequence of cerebellar disease have not been thoroughly studied. Here, we analysed structural and diffusion imaging data from cerebellar patients with chronic surgical lesions after resection for benign cerebellar tumours. We aimed to elucidate the impact of focal lesions of the cerebellum on WM integrity across the entire brain, and to investigate whether WM deficits were associated with behavioural impairment in three different motor tasks. Lesion symptom mapping analysis suggested that lesions in critical cerebellar regions were related to deficits in savings during an eyeblink conditioning task, as well as to deficits in motor action timing. Diffusion imaging analysis of cerebellar WM indicated that better behavioural performance was associated with higher fractional anisotropy (FA) in the superior cerebellar peduncle, cerebellum's main outflow path. Moreover, voxel-wise analysis revealed a global pattern of WM deficits in patients within many cerebral WM tracts critical for motor and non-motor function. Finally, we observed a positive correlation between FA and savings within cerebello-thalamo-cortical pathways in patients but not in controls, showing that saving effects partly depend on extracerebellar areas, and may be recruited for compensation. These results confirm that the cerebellum has extended connections with many cerebral areas involved in motor/cognitive functions, and the observed WM changes likely contribute to long-term clinical deficits of posterior fossa tumour survivors.

Effective connectivity differences in motor network during passive movement of paretic and non-paretic ankles in subacute stroke patients.

BACKGROUND: A better understanding of the neural changes associated with paresis in stroke patients could have important implications for therapeutic approaches. Dynamic Causal Modeling (DCM) for functional magnetic resonance imaging (fMRI) is commonly used for analyzing effective connectivity patterns of brain networks due to its significant property of modeling neural states behind fMRI signals. We applied this technique to analyze the differences between motor networks (MNW) activated by continuous passive movement (CPM) of paretic and non-paretic ankles in subacute stroke patients. This study aimed to identify CPM induced connectivity characteristics of the primary sensory area (S1) and the differences in extrinsic directed connections of the MNW and to explain the hemodynamic differences of brain regions of MNW. METHODS: For the network analysis, we used ten stroke patients' task fMRI data collected under CPMs of both ankles. Regions for the MNW, the primary motor cortex (M1), the premotor cortex (PM), the supplementary motor area (SMA) and the S1 were defined in a data-driven way, by independent component analysis. For the network analysis of both CPMs, we compared twelve models organized into two model-families, depending on the S1 connections and input stimulus modeling. Using DCM, we evaluated the extrinsic connectivity strengths and hemodynamic parameters of both stimulations of all patients. RESULTS: After a statistical comparison of the extrinsic connections and their modulations of the "best model", we concluded that three contralateral self-inhibitions (cM1, cS1 and cSMA), one contralateral inter-regional connection (cSMA→cM1), and one interhemispheric connection (cM1→iM1) were significantly different. Our research shows that hemodynamic parameters can be estimated with the Balloon model using DCM but the parameters do not change with stroke. CONCLUSIONS: Our results confirm that the DCM-based connectivity analyses combined with Bayesian model selection may be a useful technique for quantifying the alteration or differences in the characteristics of the motor network in subacute stage stroke patients and in determining the degree of MNW changes.

Brain MRI Diffusion Encoding Direction Number Affects Tract-Based Spatial Statistics Results in Multiple Sclerosis.

BACKGROUND AND PURPOSE: Diffusion tensor imaging (DTI) is a promising approach to detect the underlying brain pathology. These alterations can be seen in several diseases such as multiple sclerosis. Tract-based spatial statistics (TBSS) is an easy to use and robust way for analyzing diffusion data. The effect of acquisition parameters of DTI on TBSS has not been evaluated, especially the number of diffusion encoding directions (NDED), which is directly proportional with scan time. METHODS: We analyzed a large set of DTI data of healthy controls (N = 126) and multiple sclerosis patients (N = 78). The highest NDED (60 directions) was reduced and a tensor calculation was done separately for every subset. We calculated the mean and standard deviation of DTI parameters under the white matter mask. Moreover, the FMRIB Software Library TBSS pipeline was used on DTI images with 15, 30, 45, and 60 directions to compare differences between groups. Mean DTI parameters were compared between groups as a function of NDED. RESULTS: The mean value of FA and AD decreased with increasing number of directions. This was more pronounced in areas with smaller FA values. RD and MD were constant. The skeleton size reduced with elevating NDED along with the number of significant voxels. The TBSS analysis showed significant differences between groups throughout the majority of the skeleton and the group difference was associated with NDED. CONCLUSION: Our results suggested that results of TBSS depended on the NDED, which should be considered when comparing DTI data with varying protocols.

Pain-free resting-state functional brain connectivity predicts individual pain sensitivity.

Individual differences in pain perception are of interest in basic and clinical research as altered pain sensitivity is both a characteristic and a risk factor for many pain conditions. It is, however, unclear how individual sensitivity to pain is reflected in the pain-free resting-state brain activity and functional connectivity. Here, we identify and validate a network pattern in the pain-free resting-state functional brain connectome that is predictive of interindividual differences in pain sensitivity. Our predictive network signature allows assessing the individual sensitivity to pain without applying any painful stimulation, as might be valuable in patients where reliable behavioural pain reports cannot be obtained. Additionally, as a direct, non-invasive readout of the supraspinal neural contribution to pain sensitivity, it may have implications for translational research and the development and assessment of analgesic treatment strategies.

Purkinje cell number-correlated cerebrocerebellar circuit anomaly in the valproate model of autism.

While cerebellar alterations may play a crucial role in the development of core autism spectrum disorder (ASD) symptoms, their pathophysiology on the function of cerebrocerebellar circuit loops is largely unknown. We combined multimodal MRI (9.4 T) brain assessment of the prenatal rat valproate (VPA) model and correlated immunohistological analysis of the cerebellar Purkinje cell number to address this question. We hypothesized that a suitable functional MRI (fMRI) paradigm might show some altered activity related to disrupted cerebrocerebellar information processing. Two doses of maternal VPA (400 and 600 mg/kg, s.c.) were used. The higher VPA dose induced 3% smaller whole brain volume, the lower dose induced 2% smaller whole brain volume and additionally a focal gray matter density decrease in the cerebellum and brainstem. Increased cortical BOLD responses to whisker stimulation were detected in both VPA groups, but it was more pronounced and extended to cerebellar regions in the 400 mg/kg VPA group. Immunohistological analysis revealed a decreased number of Purkinje cells in both VPA groups. In a detailed analysis, we revealed that the Purkinje cell number interacts with the cerebral BOLD response distinctively in the two VPA groups that highlights atypical function of the cerebrocerebellar circuit loops with potential translational value as an ASD biomarker.

Probabilistic TFCE: A generalized combination of cluster size and voxel intensity to increase statistical power.

The threshold-free cluster enhancement (TFCE) approach integrates cluster information into voxel-wise statistical inference to enhance detectability of neuroimaging signal. Despite the significantly increased sensitivity, the application of TFCE is limited by several factors: (i) generalisation to data structures, like brain network connectivity data is not trivial, (ii) TFCE values are in an arbitrary unit, therefore, P-values can only be obtained by a computationally demanding permutation-test. Here, we introduce a probabilistic approach for TFCE (pTFCE), that gives a simple general framework for topology-based belief boosting. The core of pTFCE is a conditional probability, calculated based on Bayes' rule, from the probability of voxel intensity and the threshold-wise likelihood function of the measured cluster size. In this paper, we provide an estimation of these distributions based on Gaussian Random Field theory. The conditional probabilities are then aggregated across cluster-forming thresholds by a novel incremental aggregation method. pTFCE is validated on simulated and real fMRI data. The results suggest that pTFCE is more robust to various ground truth shapes and provides a stricter control over cluster "leaking" than TFCE and, in many realistic cases, further improves its sensitivity. Correction for multiple comparisons can be trivially performed on the enhanced P-values, without the need for permutation testing, thus pTFCE is well-suitable for the improvement of statistical inference in any neuroimaging workflow. Implementation of pTFCE is available at https://spisakt.github.io/pTFCE.

Stepwise occlusion of the carotid arteries of the rat: MRI assessment of the effect of donepezil and hypoperfusion-induced brain atrophy and white matter microstructural changes.

Bilateral common carotid artery occlusion (BCCAo) in the rat is a widely used animal model of vascular dementia and a valuable tool for preclinical pharmacological drug testing, although the varying degrees of acute focal ischemic lesions it induces could interfere with its translational value. Recently, a modification to the BCCAo model, the stepwise occlusion of the two carotid arteries, has been introduced. To acquire objective translatable measures, we used longitudinal multimodal magnetic resonance imaging (MRI) to assess the effects of semi-chronic (8 days) donepezil treatment in this model, with half of the Wistar rats receiving the treatment one week after the stepwise BCCAo. With an ultrahigh field MRI, we measured high-resolution anatomy, diffusion tensor imaging, cerebral blood flow measurements and functional MRI in response to whisker stimulation, to evaluate both the structural and functional effects of the donepezil treatment and stepwise BCCAo up to 5 weeks post-occlusion. While no large ischemic lesions were detected, atrophy in the striatum and in the neocortex, along with widespread white matter microstructural changes, were found. Donepezil ameliorated the transient drop in the somatosensory BOLD response in distant cortical areas, as detected 2 weeks after the occlusion but the drug had no effect on the long term structural changes. Our results demonstrate a measurable functional MRI effect of the donepezil treatment and the importance of diffusion MRI and voxel based morphometry (VBM) analysis in the translational evaluation of the rat BCCAo model.

Gray Matter Atrophy Is Primarily Related to Demyelination of Lesions in Multiple Sclerosis: A Diffusion Tensor Imaging MRI Study.

Objective: Cortical pathology, periventricular demyelination, and lesion formation in multiple sclerosis (MS) are related (Hypothesis 1). Factors in the cerebrospinal fluid close to these compartments could possibly drive the parallel processes. Alternatively, the cortical atrophy could be caused by remote axonal transection (Hypothesis 2). Since MRI can differentiate between demyelination and axon loss, we used this imaging modality to investigate the correlation between the pattern of diffusion parameter changes in the periventricular- and deep white matter and the gray matter atrophy. Methods: High-resolution T1-weighted, FLAIR, and diffusion MRI images were acquired in 52 RRMS patients and 50 healthy, age-matched controls. We used EDSS to estimate the clinical disability. We used Tract Based Spatial Statistics to compare diffusion parameters (fractional anisotropy, mean, axial, and radial diffusivity) between groups. We evaluated global brain, white, and gray matter atrophy with SIENAX. Averaged, standard diffusion parameters were calculated in four compartment: periventricular lesioned and normal appearing white matter, non-periventricular lesioned and normal appearing white matter. PLS regression was used to identify which diffusion parameter and in which compartment best predicts the brain atrophy and clinical disability. Results: In our diffusion tensor imaging study compared to controls we found extensive alterations of fractional anisotropy, mean and radial diffusivity and smaller changes of axial diffusivity (maximal p > 0.0002) in patients that suggested demyelination in the lesioned and in the normal appearing white matter. We found significant reduction in total brain, total white, and gray matter (patients: 718.764 ± 14.968, 323.237 ± 7.246, 395.527 ± 8.050 cm3, controls: 791.772 ± 22.692, 355.350 ± 10.929, 436.422 ± 12.011 cm3; mean ± SE), (p < 0.015; p < 0.0001; p < 0.009; respectively) of patients compared to controls. The PLS analysis revealed a combination of demyelination-like diffusion parameters (higher mean and radial diffusivity in patients) in the lesions and in the non-lesioned periventricular white matter, which best predicted the gray matter atrophy (p < 0.001). Similarly, EDSS was best predicted by the radial diffusivity of the lesions and the non-lesioned periventricular white matter, but axial diffusivity of the periventricular lesions also contributed significantly (p < 0.0001). Interpretation: Our investigation showed that gray matter atrophy and white matter demyelination are related in MS but white matter axonal loss does not significantly contribute to the gray matter pathology.

Population-Level Correction of Systematic Motion Artifacts in fMRI in Patients with Ischemic Stroke.

BACKGROUND: The aim of this study was to reveal potential sources of systematic motion artifacts in stroke functional magnetic resonance imaging (fMRI) focusing on those causing stimulus-correlated motion on the individual-level and separate the motion effect on the fMRI signal changing from the activation-induced alteration at population level. METHODS: Eleven ischemic stroke patients were examined by fMRI. The fMRI paradigm was based on passive ankle movement on both the healthy and the paretic leg's side. Three individual-level motion correction strategies were compared and we introduced five measures to characterize each subjects' in-scanner relative head movement. After analyzing the correlation of motion parameters and the subjects' physiological scale scores, we selected a parameter to model the motion-related artifacts in the second-level analysis. RESULTS: At first (individual) level analysis, the noise-component correction-based CompCor method provided the highest -log10(p) value of cluster-level occurrence probability at 12.4/13.6 for healthy and paretic side stimulus, respectively, with a maximal z-value of 15/16.3. Including the motion parameter at second (group) level resulted in lower cluster occurrence values at 10.9/5.55 while retaining the maximal z-value. CONCLUSIONS: We proposed a postprocessing pipeline for ischemic stroke fMRI data that combine the CompCor correction at first level with the modeling of motion effect at second-level analysis by a parameter obtained from fMRI data. Our solution is applicable for any fMRI-based stroke rehabilitation study since it does not require any MRI-compatible motion capture system and is based on commonly used methods.