The brain is the main organ of the nervous system, made up of different cell types. Two distinct regions can be identified, gray matter and white matter. The first, external, is made up of the bodies of neurons, while the second - internal - is made up of the myelin coating of their axons. There are exceptions such as deep nuclei, portions of gray matter placed inside the brain. In this thesis, attention was paid to dentate nuclei, cerebellar nuclei of extreme importance for the reconstruction of cerebro-cerebellar communication mechanisms. Using magnetic resonance diffusion-weighed images from the public database of the Human Connectome Project, a Matlab analysis pipeline was developed with the aim of reconstructing the topography of these nuclei starting from their connectivity with the different areas of the brain and cerebellum. Initially, a series of features were calculated from the images of the subjects and the standard atlases were deformed in the space of the individual subjects. The dentate masks, coming from a previous work of the neuroimaging department of Pavia, have been added to the images sectioned in the different tissues, so as to increase the sensitivity of subsequent analyzes in these small areas. 30 million neural connections were reconstructed for each subject, generating a probabilistic tractography. By selecting the sections connecting each individual parcellation of the cerebellum and the ipsilateral dentate, a membership vector was built for each voxel of the latter, whose components have as their value the number of streamlines terminated them. By associating each voxel with the most connected area, it was possible to build a connectivity map for the dentate and ipsilateral cerebellar cortex. Similarly, a map was constructed for the connection with the contralateral thalamus, therefore indirectly, with the contralateral cerebral cortex. A third approach was to use an unsupervised machine learning algorithm to generate a clustering of dentate nuclei, based solely on structural similarity, using the features calculated initially. Together, these maps represent the structural identity of the dentate nuclei, allowing to infer their role in cerebro-cerebellar communication starting from the areas to which they are connected. Connectivity-based methods have shown that approximately 50% of the volume of the toothed nuclei is associated with Crus cerebellar while the rest is mainly connected to lobules I-VI, furthermore, the mainly connected subthalamic nuclei are the ventral nucleus anterior and medial, in turn connected with the prefrontal and cortex front. These maps then revealed how dentate nuclei have an important role not only in motor and postural control, but also in a series of higher cognitive and associative functions. From the microstructural point of view, the dentate nuclei they were segmented into two distinct areas: an external lateral and a medial internal. It should be noted that these clusters overlap the parcellations well associative and motor identified with connectivity methods. In the final part of the thesis, a proof-of-concept of a filtering code written in Python developed. Despite the power of the tractography technique, it suffers from some intrinsic limitations, evident mainly in the portions of space, the fibers overlap or intersect themselves. The idea of this code is to link the tractography to physiological values, such as a constant myelin value, eliminating the streamlines that most deviate from this value. Using a cost function and a gradient descent the algorithm has been successfully tested on simple examples and visually improves the tractography using physiological information.
L’encefalo è l’organo principale del sistema nervoso, composto da diversi tipi cellulari. Si possono individuare due regioni distinte, la materia grigia e quella bianca. La prima, esterna, è composta dai corpi dei neuroni, mentre la seconda -interna- dal rivestimento mielinico dei loro assoni. Esistono eccezioni come i nuclei profondi, porzioni di materia grigia posti internamente al cervello. In questa tesi l’attenzione è stata rivolta ai nuclei dentati, nuclei cerebellari di estrema importanza per la ricostruzione dei meccanismi di comunicazione cerebro-cerebellari. Utilizzando immagini di risonanza magnetica pesate in diffusioni provenienti dal database pubblico dello Human Connectome Project è stata elaborata una pipeline di analisi di Matlab al fine di ricostruire la topografia di questi nuclei a partire dalla loro connettività con le diverse aree di cervello e cervelletto. Inizialmente, dalle immagini dei soggetti sono state calcolate una serie di features e gli atlanti standard sono stati deformati nello spazio dei singoli soggetti. Le maschere dei dentati sono state aggiunte alle immagini sezionate nei diversi tessuti, così da incrementare la sensibilità delle analisi successive in queste aree così piccole. Sono state ricostruite 30 milioni di connessioni neurali per ogni soggetto, generando una trattografia probabilistica. Selezionando i tratti congiungenti ogni singola parcellazione del cervelletto e il dentato ipsilaterale, per ogni voxel di quest’ultimo è stata costruita un vettore di appartenza, le cui componenti hanno come valore il numero di streamlines li terminate. Associando ogni voxel all’area più connessa è stato possibile costruire una mappa di connettività per il dentato e la corteccia cerebellare ipsilaterale. Analogamente è stata costruita una mappa per la connessione con il talamo controlaterale, quindi indirettamente, con la corteccia cerebrale controlaterale. Un terzo approccio è stato quello di utilizzare un algoritmo di machine learning non supervisionato per generare un clustering dei nuclei dentati, basandosi unicamente sulla similarità strutturale, utilizzando le features calcolate inizialmente. Queste mappe, insieme, rappresentano l’identità strutturale dei nuclei dentati, permettendo di inferire il loro ruolo nella comunicazione cerebro-cerebellare a partire dalle aree a cui sono connesse. I metodi basati sulla connettività hanno evidenziato che circa il 50% del volume dei nuclei dentati è associato alle Crus cerebellari mentre il restante è principalmente connesso ai lobuli I-VI, inoltre i nuclei subtalamici principalmente connessi sono il nucleo ventrale anteriore e mediale, a loro volta connessi con la corteccia prefrontale e quella frontale. Queste mappe hanno quindi rivelato come i nuclei dentati rivestano un ruolo importante non solo nel controllo motorio e posturale, ma anche in una serie di funzioni superiori di tipo cognitivo ed associativo. Dal punto di vista microstrutturale i nuclei dentati sono stati segmentati in due aree distinte: una laterale esterna e una mediale interna. Questi cluster ben si sovrappongono alle parcellazioni associative e motorie individuate con i metodi della connettività. Nella parte finale della tesi, si è sviluppato un proof-of-concept di un codice di filtraggio scritto in Python. Nonostante la potenza della tecnica della trattografia, essa soffre di alcune limitazioni intrinseche, evidenti principalmente nelle porzioni di spazio le fibre si sovrappongono o intersecano. L’idea di questo codice è quella di vincolare la trattografia a valori fisiologici, come un valore di mielina costante, eliminando i tratti che più si discostano da tale valore. Utilizzando una funzione di costo e una discesa del gradiente l’algoritmo è stato testato con successo su semplici esempi migliorando la trattografia includendo le informazioni fisiologiche.
Topografia e connettività microstrutturale dei nuclei dentati
FERRANTE, MATTEO
2019/2020
Abstract
The brain is the main organ of the nervous system, made up of different cell types. Two distinct regions can be identified, gray matter and white matter. The first, external, is made up of the bodies of neurons, while the second - internal - is made up of the myelin coating of their axons. There are exceptions such as deep nuclei, portions of gray matter placed inside the brain. In this thesis, attention was paid to dentate nuclei, cerebellar nuclei of extreme importance for the reconstruction of cerebro-cerebellar communication mechanisms. Using magnetic resonance diffusion-weighed images from the public database of the Human Connectome Project, a Matlab analysis pipeline was developed with the aim of reconstructing the topography of these nuclei starting from their connectivity with the different areas of the brain and cerebellum. Initially, a series of features were calculated from the images of the subjects and the standard atlases were deformed in the space of the individual subjects. The dentate masks, coming from a previous work of the neuroimaging department of Pavia, have been added to the images sectioned in the different tissues, so as to increase the sensitivity of subsequent analyzes in these small areas. 30 million neural connections were reconstructed for each subject, generating a probabilistic tractography. By selecting the sections connecting each individual parcellation of the cerebellum and the ipsilateral dentate, a membership vector was built for each voxel of the latter, whose components have as their value the number of streamlines terminated them. By associating each voxel with the most connected area, it was possible to build a connectivity map for the dentate and ipsilateral cerebellar cortex. Similarly, a map was constructed for the connection with the contralateral thalamus, therefore indirectly, with the contralateral cerebral cortex. A third approach was to use an unsupervised machine learning algorithm to generate a clustering of dentate nuclei, based solely on structural similarity, using the features calculated initially. Together, these maps represent the structural identity of the dentate nuclei, allowing to infer their role in cerebro-cerebellar communication starting from the areas to which they are connected. Connectivity-based methods have shown that approximately 50% of the volume of the toothed nuclei is associated with Crus cerebellar while the rest is mainly connected to lobules I-VI, furthermore, the mainly connected subthalamic nuclei are the ventral nucleus anterior and medial, in turn connected with the prefrontal and cortex front. These maps then revealed how dentate nuclei have an important role not only in motor and postural control, but also in a series of higher cognitive and associative functions. From the microstructural point of view, the dentate nuclei they were segmented into two distinct areas: an external lateral and a medial internal. It should be noted that these clusters overlap the parcellations well associative and motor identified with connectivity methods. In the final part of the thesis, a proof-of-concept of a filtering code written in Python developed. Despite the power of the tractography technique, it suffers from some intrinsic limitations, evident mainly in the portions of space, the fibers overlap or intersect themselves. The idea of this code is to link the tractography to physiological values, such as a constant myelin value, eliminating the streamlines that most deviate from this value. Using a cost function and a gradient descent the algorithm has been successfully tested on simple examples and visually improves the tractography using physiological information.È consentito all'utente scaricare e condividere i documenti disponibili a testo pieno in UNITESI UNIPV nel rispetto della licenza Creative Commons del tipo CC BY NC ND.
Per maggiori informazioni e per verifiche sull'eventuale disponibilità del file scrivere a: unitesi@unipv.it.
https://hdl.handle.net/20.500.14239/11623