Comprehensive analysis of synthetic learning applied to neonatal brain MRI segmentation.

Brain segmentation from neonatal MRI images is a very challenging task due to large changes in the shape of cerebral structures and variations in signal intensities reflecting the gestational process. In this context, there is a clear need for segmentation techniques that are robust to variations in image contrast and to the spatial configuration of anatomical structures. In this work, we evaluate the potential of synthetic learning, a contrast-independent model trained using synthetic images generated from the ground truth labels of very few subjects. We base our experiments on the dataset released by the developmental Human Connectome Project, for which high-quality images are available for more than 700 babies aged between 26 and 45 weeks postconception. First, we confirm the impressive performance of a standard UNet trained on a few volumes, but also confirm that such models learn intensity-related features specific to the training domain. We then confirm the robustness of the synthetic learning approach to variations in image contrast. However, we observe a clear influence of the age of the baby on the predictions. We improve the performance of this model by enriching the synthetic training set with realistic motion artifacts and over-segmentation of the white matter. Based on extensive visual assessment, we argue that the better performance of the model trained on real T2w data may be due to systematic errors in the ground truth. We propose an original experiment allowing us to show that learning from real data will reproduce any systematic bias affecting the training set, while synthetic models can avoid this limitation. Overall, our experiments confirm that synthetic learning is an effective solution for segmenting neonatal brain MRI. Our adapted synthetic learning approach combines key features that will be instrumental for large multisite studies and clinical applications.

Dopamine denervation in the functional territories of the striatum: a new MR and atlas-based 123I-FP-CIT SPECT quantification method.

Current quantification methods of 123I-FP-CIT SPECT rely on anatomical parcellation of the striatum. We propose here to implement a new method based on MRI segmentation and functional atlas of the basal ganglia (MR-ATLAS) that could provide a reliable quantification within the sensorimotor, associative, and limbic territories of the striatum. Patients with Parkinson's disease (PD), idiopathic rapid eye movement sleep behavioral disorder (iRBD), and healthy controls underwent 123I-FP-CIT SPECT, MRI, motor, and cognitive assessments. SPECT data were corrected for partial volume effects and registered to a functional atlas of the striatum to allow quantification in every functional region of the striatum (nucleus accumbens, limbic, associative, and sensorimotor parts of the striatum). The MR-ATLAS quantification method is proved to be reliable in every territory of the striatum. In addition, good correlations were found between cognitive dysexecutive tests and the binding within the functional (limbic) territories of the striatum using the MR-ATLAS method, slightly better than correlations found using the anatomical quantification method. This new MR-ATLAS method provides a robust and useful tool for studying the dopaminergic system in PD, particularly with respect to cognitive functions. It may also be relevant to further unravel the relationship between dopaminergic denervation and cognitive or behavioral symptoms.

Comparative Study of MRI Biomarkers in the Substantia Nigra to Discriminate Idiopathic Parkinson Disease.

BACKGROUND AND PURPOSE: Several new MR imaging techniques have shown promising results in patients with Parkinson disease; however, the comparative diagnostic values of these measures at the individual level remain unclear. Our aim was to compare the diagnostic value of MR imaging biomarkers of substantia nigra damage for distinguishing patients with Parkinson disease from healthy volunteers. MATERIALS AND METHODS: Thirty-six patients and 20 healthy volunteers were prospectively included. The MR imaging protocol at 3T included 3D T2-weighted and T1-weighted neuromelanin-sensitive images, diffusion tensor images, and R2* mapping. T2* high-resolution images were also acquired at 7T to evaluate the dorsal nigral hyperintensity sign. Quantitative analysis was performed using ROIs in the substantia nigra drawn manually around the area of high signal intensity on neuromelanin-sensitive images and T2-weighted images. Visual analysis of the substantia nigra neuromelanin-sensitive signal intensity and the dorsolateral nigral hyperintensity on T2* images was performed. RESULTS: There was a significant decrease in the neuromelanin-sensitive volume and signal intensity in patients with Parkinson disease. There was also a significant decrease in fractional anisotropy and an increase in mean, axial, and radial diffusivity in the neuromelanin-sensitive substantia nigra at 3T and a decrease in substantia nigra volume on T2* images. The combination of substantia nigra volume, signal intensity, and fractional anisotropy in the neuromelanin-sensitive substantia nigra allowed excellent diagnostic accuracy (0.93). Visual assessment of both substantia nigra dorsolateral hyperintensity and neuromelanin-sensitive images had good diagnostic accuracy (0.91 and 0.86, respectively). CONCLUSIONS: The combination of neuromelanin signal and volume changes with fractional anisotropy measurements in the substantia nigra showed excellent diagnostic accuracy. Moreover, the high diagnostic accuracy of visual assessment of substantia nigra changes using dorsolateral hyperintensity analysis or neuromelanin-sensitive signal changes indicates that these techniques are promising for clinical practice.

White matter predicts functional connectivity in premanifest Huntington's disease.

OBJECTIVES: The distribution of pathology in neurodegenerative disease can be predicted by the organizational characteristics of white matter in healthy brains. However, we have very little evidence for the impact these pathological changes have on brain function. Understanding any such link between structure and function is critical for understanding how underlying brain pathology influences the progressive behavioral changes associated with neurodegeneration. Here, we demonstrate such a link between structure and function in individuals with premanifest Huntington's. METHODS: Using diffusion tractography and resting state functional magnetic resonance imaging to characterize white matter organization and functional connectivity, we investigate whether characteristic patterns of white matter organization in the healthy human brain shape the changes in functional coupling between brain regions in premanifest Huntington's disease. RESULTS: We find changes in functional connectivity in premanifest Huntington's disease that link directly to underlying patterns of white matter organization in healthy brains. Specifically, brain areas with strong structural connectivity show decreases in functional connectivity in premanifest Huntington's disease relative to controls, while regions with weak structural connectivity show increases in functional connectivity. Furthermore, we identify a pattern of dissociation in the strongest functional connections between anterior and posterior brain regions such that anterior functional connectivity increases in strength in premanifest Huntington's disease, while posterior functional connectivity decreases. INTERPRETATION: Our findings demonstrate that organizational principles of white matter underlie changes in functional connectivity in premanifest Huntington's disease. Furthermore, we demonstrate functional antero-posterior dissociation that is in keeping with the caudo-rostral gradient of striatal pathology in HD.

Reproducibility of R2 * and quantitative susceptibility mapping (QSM) reconstruction methods in the basal ganglia of healthy subjects.

The basal ganglia are key structures for motor, cognitive and behavioral functions. They undergo several changes with aging and disease, such as Parkinson's or Huntington's disease, for example. Iron accumulation in basal ganglia is often related to these diseases, which is conventionally monitored by the transverse relaxation rate (R2 *). Quantitative susceptibility mapping (QSM) is a novel contrast mechanism in MRI produced by adding information taken from the phase of the MR signal to its magnitude. It has been shown to be more sensitive to subtle changes in Parkinson's disease. In order to be applied widely to various pathologies, its reproducibility must be evaluated in order to assess intra-subject variability and to disseminate into clinical and pharmaceutical studies. In this work, we studied the reproducibility and sensitivity of several QSM techniques. Fourteen subjects were scanned four times, and QSM and R2 * images were reconstructed and registered. An atlas of the basal ganglia was used to automatically define regions of interest. We found that QSM measurements are indeed reproducible in the basal ganglia of healthy subjects and can be widely used as a replacement for R2 * mapping in iron-rich regions. This reproducibility study could lead to several lines of research in relaxometry and susceptibility measurements, in vivo iron load evaluation as well as pharmacological assessment and biomarker development. Copyright © 2016 John Wiley & Sons, Ltd.

Increased cortico-striatal connectivity during motor practice contributes to the consolidation of motor memory in writer's cramp patients.

Sensorimotor representations of movements are created in the sensorimotor network through repeated practice to support successful and effortless performance. Writer's cramp (WC) is a disorder acquired through extensive practice of finger movements, and it is likely associated with the abnormal acquisition of sensorimotor representations. We investigated (i) the activation and connectivity changes in the brain network supporting the acquisition of sensorimotor representations of finger sequences in patients with WC and (ii) the link between these changes and consolidation of motor performance 24 h after the initial practice. Twenty-two patients with WC and 22 age-matched healthy volunteers practiced a complex sequence with the right (pathological) hand during functional MRI recording. Speed and accuracy were measured immediately before and after practice (day 1) and 24 h after practice (day 2). The two groups reached equivalent motor performance on day 1 and day 2. During motor practice, patients with WC had (i) reduced hippocampal activation and hippocampal-striatal functional connectivity; and (ii) overactivation of premotor-striatal areas, whose connectivity correlated with motor performance after consolidation. These results suggest that patients with WC use alternative networks to reach equiperformance in the acquisition of new motor memories.

Broca's area damage is necessary but not sufficient to induce after-effects of cathodal tDCS on the unaffected hemisphere in post-stroke aphasia.

BACKGROUND: The inter-individual variability of behavioral effects after tDCS applied to the unaffected right hemisphere in stroke may be related to factors such as the lesion location. OBJECTIVE/HYPOTHESIS: We investigated the effect of left Broca's area (BA) damage on picture naming in aphasic patients after cathodal tDCS applied over the right BA. METHODS: We conducted a study using pre-interventional diffusion and resting state functional MRI (rsfMRI) and two cross-over tDCS sessions (TYPE: sham and cathodal) over the right homologous BA in aphasic stroke patients with ischemic lesions involving the left BA (BA+) or other left brain areas (BA-). Picture naming accuracy was assessed after each session. Inter-hemispheric (IH) functional balance was investigated via rsfMRI connectivity maps using the right BA as a seed. Probabilistic tractography was used to study the integrity of language white matter pathways. RESULTS: tDCS had different effects on picture naming accuracy in BA+ and BA- patients (TYPE × GROUP interaction, F(1,19): 4.6, P: 0.04). All BA- patients except one did not respond to tDCS and demonstrated normal IH balance between the right and left BA when compared to healthy subjects. BA+ patients were improved by tDCS in 36% and had decreased level of functional IH balance. Improvement in picture naming after cathodal tDCS was associated with the integrity of the arcuate fasciculus in BA+ patients. CONCLUSIONS: Behavioral effects of cathodal tDCS on the unaffected right hemisphere differ depending on whether BA and the arcuate fasciculus are damaged. Therefore, IH imbalance could be a direct consequence of anatomical lesions.

Connectivity between right inferior frontal gyrus and supplementary motor area predicts after-effects of right frontal cathodal tDCS on picture naming speed.

BACKGROUND: Cathodal transcranial direct current stimulation (tDCS) of the right frontal cortex improves language abilities in post-stroke aphasic patients. Yet little is known about the effects of right frontal cathodal tDCS on normal language function. OBJECTIVE/HYPOTHESIS: To explore the cathodal tDCS effects of the right-hemispheric homologue of Broca's area on picture naming in healthy individuals. We hypothesized that cathodal tDCS improves picture naming and that this effect is determined by the anatomical and functional connectivity of the targeted region. METHODS: Cathodal and sham tDCS were applied to the right inferior frontal gyrus in 24 healthy subjects before a picture-naming task. All participants were studied with magnetic resonance imaging at pre-interventional baseline. Probabilistic tractography and dynamic causal modeling of functional brain activity during a word repetition task were applied to characterize anatomical and functional connectivity. RESULTS: Subjects named pictures faster after cathodal relative to sham tDCS. The accelerating effect of tDCS was explained by a reduced frequency of very slow responses. tDCS-induced acceleration of picture naming correlated with larger volumes of the tract connecting the right Broca's area and the supplementary motor area (SMA) and greater functional coupling from the right SMA to the right Broca's area. CONCLUSIONS: The results support the notion that the after-effects of tDCS on brain function are at least in part determined by the anatomical and functional connectivity of the targeted region.

Early ADC changes in motor structures predict outcome of acute stroke better than lesion volume.

OBJECTIVES: The lesion volume assessed from diffusion-weighted imaging (DWI) within the first six hours to first week following stroke onset has been proposed as a predictor of functional outcome in clinical studies. However, the prediction accuracy decreases when the DWI lesion volume is measured during the earliest stages of patient evaluation. In this study, our hypothesis was that the combination of lesion location (motor-related regions) and diffusivity measures (such as Apparent Diffusion Coefficient [ADC]) at the acute stage of stroke predict clinical outcome. PATIENTS AND METHODS: Seventy-nine consecutive acute carotid territory stroke patients (median age: 62 years) were included in the study and outcome at three months was assessed using the modified Rankin scale (good outcome: mRS 0-2; poor outcome: mRS 3-5). DWI was acquired within the first six hours of stroke onset (H2) and the following day (D1). Apparent Diffusion Coefficient (ADC) values were measured in the corticospinal tract (CST), the primary motor cortex (M1), the supplementary motor area (SMA), the putamen in the affected hemisphere, and in the contralateral cerebellum to predict stroke outcome. RESULTS: Prediction of poor vs. good outcome at the individual level at H2 (D1, respectively) was achieved with 74% accuracy, 95%CI: 53-89% (75%, 95% CI: 61-89%, respectively) when patients were classified from ADC values measured in the putamen and CST. Prediction accuracy from DWI volumes reached only 62% (95%CI: 42-79%) at H2 and 69% (95%CI: 50-85%) at D1. CONCLUSION: We therefore show that measures of ADC at the acute stage in deeper motor structures (putamen and CST) are better predictors of stroke outcome than DWI lesion volume.

Individual differences in prefrontal cortical activation on the Tower of London planning task: implication for effortful processing.

Solving challenging ('effortful') problems is known to involve the dorsal and dorsolateral prefrontal cortex in normal volunteers, although there is considerable individual variation. In this functional magnetic resonance imaging study, we show that healthy subjects with different levels of performance in the Tower of London planning task exhibit different patterns of brain activation. All subjects exhibited significant bilateral activation in the dorsolateral prefrontal cortex, the anterior and posterior cingulate areas and the parietal cortex. However, 'standard performers' (performance < 70% correct) and 'superior performers' (performance >70% correct) differed in the patterns of activation exhibited. Superior performers showed a significantly more spatially extended activation in the left dorsolateral prefrontal cortex than did standard performers, whereas the latter group tended to show increased activation of the anterior cingulate region.

Modelling of the coupling between brain electrical activity and metabolism.

In order to make an attempt at grouping the various aspects of brain functional imaging (fMRI, MRS, EEG-MEG, ...) within a coherent frame, we implemented a model consisting of a system of differential equations, that includes: (1) sodium membrane transport, (2) Na/K ATPase, (3) neuronal energy metabolism (i.e. glycolysis, buffering effect of phosphocreatine, and mitochondrial respiration), (4) blood-brain barrier exchanges and (5) brain hemodynamics, all the processes which are involved in the activation of brain areas. We assumed that the correlation between brain activation and metabolism could be due to either changes in the concentrations of ATP and ADP following activation of Na/K ATPase that result from the changes in ion concentrations, or the involvement, in different phases of metabolism, of a second messenger such as calcium. In this article, we show how this type of model enables interpretation of MRS and fMRI published data that were obtained during prolonged stimulations.

The basins of attraction of a new Hopfield learning rule.

The nature of the basins of attraction of a Hopfield network is as important as the capacity. Here a new learning rule is re-introduced. This learning rule has a higher capacity than that of the Hebb rule, and still keeps important functionality, such as incrementality and locality, which the pseudo-inverse lacks. However the basins of attraction of the fixed points of this learning rule have not yet been studied. Three important characteristics of basins of attraction are considered: indirect and direct basins of attraction, distribution of sizes of basins of attraction and the shape of the basins of attraction. The results for the new learning rule are compared with those of the Hebb rule. The size of direct and indirect basins of attractions are generally larger for the new rule than for the Hebb rule, the distribution of sizes is more even, and the shape of the basins more round.