Dependence of Connectivity on the Logarithm of Geometric Distance in Brain Networks

Machine Learning Recognizes Stages of Parkinson's Disease Using Magnetic Resonance Imaging

Neurodegenerative diseases (NDs), such as Alzheimer's disease (AD) and Parkinson's disease (PD), are debilitating conditions that affect millions worldwide, and the number of cases is expected to rise significantly in the coming years. Because early detection is crucial for effective intervention strategies, this study investigates whether the structural analysis of selected brain regions, including volumes and their spatial relationships obtained from regular T1-weighted MRI scans (N = 168, PPMI database), can model stages of PD using standard machine learning (ML) techniques. Thus, diverse ML models, including Logistic Regression, Random Forest, Support Vector Classifier, and Rough Sets, were trained and evaluated. Models used volumes, Euclidean, and Cosine distances of subcortical brain structures relative to the thalamus to differentiate among control (HC), prodromal (PR), and PD groups. Based on three separate experiments, the Logistic Regression approach was optimal, providing low feature complexity and strong predictive performance (accuracy: 85%, precision: 88%, recall: 85%) in PD-stage recognition. Using interpretable metrics, such as the volume- and centroid-based spatial distances, models achieved high diagnostic accuracy, presenting a promising framework for early-stage PD identification based on MRI scans.

Identification of miRNAs regulating MAPT expression and their analysis in plasma of patients with dementia

Background

Dementia is one of the most common diseases in elderly people and hundreds of thousand new cases per year of Alzheimer's disease (AD) are estimated. While the recent decade has seen significant advances in the development of novel biomarkers to identify dementias at their early stage, a great effort has been recently made to identify biomarkers able to improve differential diagnosis. However, only few potential candidates, mainly detectable in cerebrospinal fluid (CSF), have been described so far.

Methods

We searched for miRNAs regulating MAPT translation. We employed a capture technology able to find the miRNAs directly bound to the MAPT transcript in cell lines. Afterwards, we evaluated the levels of these miRNAs in plasma samples from FTD (n = 42) and AD patients (n = 33) and relative healthy controls (HCs) (n = 42) by using qRT-PCR.

Results

Firstly, we found all miRNAs that interact with the MAPT transcript. Ten miRNAs have been selected to verify their effect on Tau levels increasing or reducing miRNA levels by using cell transfections with plasmids expressing the miRNAs genes or LNA antagomiRs. Following the results obtained, miR-92a-3p, miR-320a and miR-320b were selected to analyse their levels in plasma samples of patients with FTD and AD respect to HCs. The analysis showed that the miR-92a-1-3p was under-expressed in both AD and FTD compared to HCs. Moreover, miR-320a was upregulated in FTD vs. AD patients, particularly in men when we stratified by sex. Respect to HC, the only difference is showed in men with AD who have reduced levels of this miRNA. Instead, miR-320b is up-regulated in both dementias, but only patients with FTD maintain this trend in both genders.

Conclusions

Our results seem to identify miR-92a-3p and miR-320a as possible good biomarkers to discriminate AD from HC, while miR-320b to discriminate FTD from HC, particularly in males. Combining three miRNAs improves the accuracy only in females, particularly for differential diagnosis (FTD vs. AD) and to distinguish FTD from HC.

Non-invasive mapping of epileptogenic networks predicts surgical outcome

Epilepsy is increasingly considered a disorder of brain networks. Studying these networks with functional connectivity can help identify hubs that facilitate the spread of epileptiform activity. Surgical resection of these hubs may lead patients who suffer from drug-resistant epilepsy to seizure freedom. Here, we aim to map non-invasively epileptogenic networks, through the virtual implantation of sensors estimated with electric and magnetic source imaging, in patients with drug-resistant epilepsy. We hypothesize that highly connected hubs identified non-invasively with source imaging can predict the epileptogenic zone and the surgical outcome better than spikes localized with conventional source localization methods (dipoles). We retrospectively analysed simultaneous high-density electroencephalography (EEG) and magnetoencephalography data recorded from 37 children and young adults with drug-resistant epilepsy who underwent neurosurgery. Using source imaging, we estimated virtual sensors at locations where intracranial EEG contacts were placed. On data with and without spikes, we computed undirected functional connectivity between sensors/contacts using amplitude envelope correlation and phase locking value for physiologically relevant frequency bands. From each functional connectivity matrix, we generated an undirected network containing the strongest connections within sensors/contacts using the minimum spanning tree. For each sensor/contact, we computed graph centrality measures. We compared functional connectivity and their derived graph centrality of sensors/contacts inside resection for good (n = 22, ILAE I) and poor (n = 15, ILAE II-VI) outcome patients, tested their ability to predict the epileptogenic zone in good-outcome patients, examined the association between highly connected hubs removal and surgical outcome and performed leave-one-out cross-validation to support their prognostic value. We also compared the predictive values of functional connectivity with those of dipoles. Finally, we tested the reliability of virtual sensor measures via Spearman's correlation with intracranial EEG at population- and patient-level. We observed higher functional connectivity inside than outside resection (P < 0.05, Wilcoxon signed-rank test) for good-outcome patients, on data with and without spikes across different bands for intracranial EEG and electric/magnetic source imaging and few differences for poor-outcome patients. These functional connectivity measures were predictive of both the epileptogenic zone and outcome (positive and negative predictive values ≥55%, validated using leave-one-out cross-validation) outperforming dipoles on spikes. Significant correlations were found between source imaging and intracranial EEG measures (0.4 ≤ rho ≤ 0.9, P < 0.05). Our findings suggest that virtual implantation of sensors through source imaging can non-invasively identify highly connected hubs in patients with drug-resistant epilepsy, even in the absence of frank epileptiform activity. Surgical resection of these hubs predicts outcome better than dipoles.

Power shift and connectivity changes in healthy aging during resting-state EEG

The neural activity of human brain changes in healthy individuals during aging. The most frequent variation in patterns of neural activity are a shift from posterior to anterior areas and a reduced asymmetry between hemispheres. These patterns are typically observed during task execution and by using functional magnetic resonance imaging data. In the present study we investigated whether analogous effects can also be detected during rest and by means of source-space time series reconstructed from electroencephalographic recordings. By analyzing oscillatory power distribution across the brain we indeed found a shift from posterior to anterior areas in older adults. We additionally examined this shift by evaluating connectivity and its changes with age. The findings indicated that inter-area connections among frontal, parietal and temporal areas were strengthened in older individuals. A more complex pattern was shown in intra-area connections, where age-related activity was enhanced in parietal and temporal areas, and reduced in frontal areas. Finally, the resulting network exhibits a loss of modularity with age. Overall, the results extend to resting-state condition the evidence of an age-related shift of brain activity from posterior to anterior areas, thus suggesting that this shift is a general feature of the aging brain rather than being task-specific. In addition, the connectivity results provide new information on the reorganization of resting-state brain activity in aging.

Relationship between mutual information and cross-correlation time scale of observability as measures of connectivity strength

The task of identifying and characterizing network structures out of experimentally observed time series is tackled by implementing different solutions, ranging from entropy-based techniques to the evaluation of the significance of observed correlation estimators. Among the metrics that belong to the first class, mutual information is of major importance due to the relative simplicity of implementation and its relying on the crucial concept of entropy. With regard to the second class, a method that allows us to assess the connectivity strength of a link in terms of a time scale of its observability via the significance estimate of measured cross correlation was recently shown to provide a reliable tool to study network structures. In this paper, we investigate the relationship between this last metric and mutual information by simultaneously assessing both metrics on large sets of data extracted from three experimental contexts, human brain magnetoencephalography, human brain electroencephalography, and surface wind measurements carried out on a small regional scale, as well as on simulated coupled, auto-regressive processes. We show that the relationship is well described by a power law and provide a theoretical explanation based on a simple noise and signal model. Besides further upholding the reliability of cross-correlation time scale of observability, the results show that the combined use of this metric and mutual information can be used as a valuable tool to identify and characterize connectivity links in a wide range of experimental contexts.