1,720,970 research outputs found
Studying brain connectivity: a new multimodal approach for structure and function integration
Full text linkIl cervello è un sistema che integra organizzazioni anatomiche e funzionali. Negli ultimi dieci anni, la comunità neuroscientifica si è posta la domanda sulla relazione struttura-funzione. Essa può essere esplorata attraverso lo studio della connettività. Nello specifico, la connettività strutturale può essere definita dal segnale di risonanza magnetica pesato in diffusione seguito dalla computazione della trattografia; mentre la correlazione funzionale del cervello può essere calcolata a partire da diversi segnali, come la risonanza magnetica funzionale o l’elettro-/magneto-encefalografia, che consente la cattura del segnale di attivazione cerebrale a una risoluzione temporale più elevata. Recentemente, la relazione struttura-funzione è stata esplorata utilizzando strumenti di elaborazione del segnale sui grafi, che estendono e generalizzano le operazioni di elaborazione del segnale ai grafi. In specifico, alcuni studi utilizzano la trasformata di Fourier applicata alla connettività strutturale per misurare la decomposizione del segnale funzionale in porzioni che si allineano (“aligned”) e non si allineano (“liberal”) con la sottostante rete di materia bianca. Il relativo allineamento funzionale con l’anatomia è stato associato alla flessibilità cognitiva, sottolineando forti allineamenti di attività corticali, e suggerendo che i sistemi sottocorticali contengono più segnali liberi rispetto alla corteccia. Queste relazioni multimodali non sono, però, ancora chiare per segnali con elevata risoluzione temporale, oltre ad essere ristretti a specifiche zone cerebrali. Oltretutto, al giorno d'oggi la ricostruzione della trattografia è ancora un argomento impegnativo, soprattutto se utilizzata per l'estrazione della connettività strutturale. Nel corso dell'ultimo decennio si è vista una proliferazione di nuovi modelli per ricostruire la trattografia, ma il loro conseguente effetto sullo strumento di connettività non è ancora chiaro. In questa tesi, ho districato i dubbi sulla variabilità dei trattogrammi derivati da diversi metodi di trattografia, confrontandoli con un paradigma di test-retest, che consente di definire la specificità e la sensibilità di ciascun modello. Ho cercato di trovare un compromesso tra queste, per definire un miglior metodo trattografico. Inoltre, ho affrontato il problema dei grafi pesati confrontando alcune possibili stime, evidenziando la sufficienza della connettività binaria e la potenza delle proprietà microstrutturali di nuova generazione nelle applicazioni cliniche. Qui, ho sviluppato un modello di proiezione che consente l'uso dei filtri aligned e liberal per i segnali di encefalografia. Il modello estende i vincoli strutturali per considerare le connessioni indirette, che recentemente si sono dimostrate utili nella relazione struttura-funzione. I risultati preliminari del nuovo modello indicano un’implicazione dinamica di momenti più aligned e momenti più liberal, evidenziando le fluttuazioni presenti nello stato di riposo. Inoltre, viene presentata una relazione specifica di periodi più allineati e liberali per il paradigma motorio. Questo modello apre la prospettiva alla definizione di nuovi biomarcatori. Considerando che l’encefalografia è spesso usata nelle applicazioni cliniche, questa integrazione multimodale applicata su dati di Parkinson o di ictus potrebbe combinare le informazioni dei cambiamenti strutturali e funzionali nelle connessioni cerebrali, che al momento sono state dimostrate individualmente.The brain is a complex system of which anatomical and functional organization is both segregated and integrated. A longstanding question for the neuroscience community has been to elucidate the mutual influences between structure and function. To that aim, first, structural and functional connectivity need to be explored individually. Structural connectivity can be measured by the Diffusion Magnetic Resonance signal followed by successive computational steps up to virtual tractography. Functional connectivity can be established by correlation between the brain activity time courses measured by different modalities, such as functional Magnetic Resonance Imaging or Electro/Magneto Encephalography. Recently, the Graph Signal Processing (GSP) framework has provided a new way to jointly analyse structure and function. In particular, this framework extends and generalizes many classical signal-processing operations to graphs (e.g., spectral analysis, filtering, and so on). The graph here is built by the structural connectome; i.e., the anatomical backbone of the brain where nodes represent brain regions and edge weights strength of structural connectivity. The functional signals are considered as time-dependent graph signals; i.e., measures associated to the nodes of the graph. The concept of the Graph Fourier Transform then allows decomposing regional functional signals into, on one side, a portion that strongly aligned with the underlying structural network (“aligned"), and, on the other side, a portion that is not well aligned with structure (“liberal"). The proportion of aligned-vs-liberal energy in functional signals has been associated with cognitive flexibility. However, the interpretation of these multimodal relationships is still limited and unexplored for higher temporal resolution functional signals such as M/EEG. Moreover, the construction of the structural connectome itself using tractography is still a challenging topic, for which, in the last decade, many new advanced models were proposed, but their impact on the connectome remains unclear. In the first part of this thesis, I disentangled the variability of tractograms derived from different tractography methods, comparing them with a test-retest paradigm, which allows to define specificity and sensitivity of each model. I want to find the best trade-off between specificity and sensitivity to define the best model that can be deployed for analysis of functional signals. Moreover, I addressed the issue of weighing the graph comparing few estimates, highlighting the sufficiency of binary connectivity, and the power of the latest-generation microstructural properties in clinical applications. In the second part, I developed a GSP method that allows applying the aligned and liberal filters to M/EEG signals. The model extends the structural constraints to consider indirect connections, which recently demonstrated to be powerful in the structure/function link. I then show that it is possible to identify dynamic changes in aligned-vs-liberal energy, highlighting fluctuations present motor task and resting state. This model opens the perspective of novel biomarkers. Indeed, M/EEG are often used in clinical applications; e.g., multimodal integration in data from Parkinson’s disease or stroke could combine changes of both structural and functional connectivity
Analysis of GFA changes along the subcortical motor network after stroke
No full textIn this work we aim at investigating the 3D Simple Harmonic Oscillator based Reconstruction and Estimation3 (3D-SHORE) derived numerical biomarkers for tractometry. In particular, we target the subcortical motor network (SC-MN) of a healthy subject. Using diffusion spectrum imaging (DSI) we reconstructed the SC-MN and compared the resulting information about white matter (WM) density and structure with that provided by Generalized Fractional Anisotropy (GFA) and Magnetization Transfer Ratio (MTR) imaging. The SC-MN gathers the connections between the cortical motor area, the basal ganglia and the thalamus, and it essentially consists of three major subcortical networks (Figure 1): i) the sensory-motor sub-loop (primary and sensory motor areas - putamen- globus pallidus - ventral lateral thalamic nucleus - motor cortex), ii) the premotor sub-loop (premotor dorsal and ventral areas (dPM, vPM) - caudate nucleus - putamen, globus pallidus –ventral anterior thalamic nuleus - premotor cortex iii) the supplementary motor area (SMA) sub-loop (SMA - putamen and caudatus - globus pallidus - ventral anterior and the ventro - lateral thalamic nuclei - SMA)
Can we trust structural connectivity?
No full textStructural connectivity models result from a complex processing chain involving many different steps, each having an impact on the reliability of the final measures. One of the hottest questions in the state-of-the-art is thus “To which extent can we trust the structural connectome?”. In this work, we tackled this issue by focusing on the typical processing pipeline and investigating the impact of the main involved steps. MRTrix CSD-based probabilistic tractography provided the highest stability across subjects and MRTrix reached the largest distance with respect to other softwares in both individual subject and group analysis
Characterization of diffusion MRI signal non Gaussianity using MAPMRI
No full textPURPOSE Diffusion in restricted media, such as the neuron axons, is nonGaussian 1. Mean Apparent Propagator Magnetic Resonance Imaging (MAPMRI 2,3 ) is a reconstruction model for diffusion MRI which is able to estimating the nonGaussianity 2 (NG) of the diffusion signal. This study aims at investigating the minimum requirements of a diffusion weighted acquisition, in terms of Signal to Noise Ratio (SNR) and maximum bvalue, for MAPMRI to capture the NG of the signal. METHODS We used Camino (http://camino.cs.ucl.ac.uk/) MonteCarlo to simulate the diffusion signal inside a pack of parallel cylinders oriented along the z axis (100000 spins 4 , 1000 timesteps 4, radius 0.5μm, and 0.1μm of space between the cylinders). We acquired the diffusion signal using 10 different sampling schemes, Δ=57.9ms, δ=13.8ms, TE=91.3ms, and considering 10 b 0, 60 directions at bvalue= 700s/mm 2 ( first shell), and 60 directions at bvalues 1000, 2000, 3000, 4000, 5000, 6000, 7000, 8000, 9000, and 10000 s/mm 2, r espectively, per sampling scheme (second shell). We fit both Diffusion Tensor Imaging 5 (DTI) and MAPMRI to the simulated data and the NG index was computed using MAPMRI. Successively, the Mean Squared Error (MSE) between ground truth and fitted signal was calculated for both DTI and MAPMRI. The same analyses were performed also adding Rician Noise to the diffusion signal with SNR=[40, 30, 20] considering 100 different instances of noise per sampling scheme. The MSE in this case was calculated with respect to the noiseless ground truth signal
Tractometry of the subcortical motor network using SHORE- based indices
No full textIntroduction. In this work we investigated the 3D-SHORE numerical indices for quantitative tractography in the sub-cortical
motor network (SC-MN). Using DSI we reconstructed the network connections and compared the outcomes about white
matter (WM) density and structure to those provided by GFA and MTR.
Methods. Ten healthy subjects (age 56.1±17.8 years old, mean±SD) went through a DSI scan twice one month apart (±
1 week, tp1c and tp2c, see [1] for more details). The Ensemble Average Propagator (EAP) was reconstructed using the
SHORE model and the orientational (ODF) and microstructural indices were derived including Return to zero (RTOP),
Return to axis (RTAP) and Return to plan (RTPP) probability and propagator anisotropy (PA). These provide an estimation
of the mean pore geometry (volume, cross-section, length, diameter), irrespectively of the pore shape. SHORE indices
were extracted for each fiber bundle
SHORE based microstructural indices: do they tell us more?
No full textDiffusion weighted magnetic resonance signals convey informationabout tissue microstructure and cytoarchitecture. Inthe last years, many models have been proposed for recoveringthe diffusion signal and extracting information to constitutenew families of microstructural indices. Here we focuson three leading diffusion MRI models: NODDI (NeuriteOrientation Dispersion and Density Imaging), 3D-SHORE(3D Simple Harmonic Oscillator-based Reconstruction andEstimation) and its formulation in the Cartesian space, theMAPMRI (Mean Apparent Propagator MRI) and analyze theinformation conveyed by the respective set of indices basedon information-theoretic measures. This will allow to objectivelyassess the ability of each index of capturing microstructuralfeatures and thus to shade light on their exploitability indiscriminative tasks. To this end, the microstructural descriptorsare treated as machine learning features and analyzedvia information-theoretic methods using in-vivo data. Resultsshow that 3D-SHORE and MAPMARI models provide indiceswith the highest relevance and that the combination ofindices from all models may provide the best ensemble offeatures for classification
Cortico-Subcortical motor network integrity relates to functional recovery after stroke
No full textIn this work, we investigated whether the structural properties of cortico-subcortical (CS) motor circuits are related to motor outcome after stroke. To do this, we acquired Diffusion Spectrum Imaging data in 10 stroke patients at 3 time points after stroke. We then performed tractographic reconstruction and estimated a number of microstructural indices, derived from the 3D Simple Harmonic Oscillator based Reconstruction and Estimation (SHORE) model, in the cortico-subcortical motor fiber tracts. Linear regression analysis showed that SHORE metrics of thalamo-cortical and intrastriatal connections in the first week after stroke are strongly related to stroke recovery at 6 months follow-up
Graph-based analysis of the structural connectivity network modulation in stroke patients
No full textFiber bundles in the brain may be reconstructed from diffusion magnetic resonance imaging data. Structural brain networks may be modeled using connectomics techniques[sup][1][/sup]. Network analyses suggest that hierarchical modular brain networks are particularly suited to facilitate local neuronal operations and the global integration of segregated functions[sup][2][/sup]. Neurological pathologies impacted to structural networks, such that the differential characterization of their topology in pathological versus healthy conditions could lead to improved diagnosis and follow-up as well as the definition of new numerical biomarkers[sup][3,4][/sup]. In this work we investigated structural connectivity, performed by Diffusion Spectrum Imaging (DSI) followed by streamline tractography, in patients affected by stroke. Connectivity matrices were extracted for microstructural indices derived from the 3D Simple Harmonic Oscillator based Reconstruction and Estimation (SHORE) model[sup][5][/sup] and for magnetization transfer ratio (MTR) and graph models were derived
Exploiting Machine Learning Principles for Assessing the Fingerprinting Potential of Connectivity Features
No full textTo which extent connectivity measures are able to characterize subjective features? The pipeline leading from the signal acquisition to the connectivity matrix allows numerous degrees of freedom each having an impact on the nal result. In this paper, we investigated the sensitivity and specicity of the connectivity models within a machine learning framework through the assessment of the detectability of repeated measures of the same subject versus other subjects. Two ber Orientation Distribution Function (fODF) reconstruction methods, one of which rstly proposed in this paper, three tractography algorithms and four connectivity features were considered and performance was expressed in terms of Area Under the Curve of the test-retest recognition task. Results suggest that there is a trade-o between the selectivity of the fODF reconstruction methods and the conservativeness of the ber tracking algorithms across all microstructural indices. The best solution was provided by using an high angular resolution fODF estimation method and the most restrictive deterministic tractography algorithm
- …
