Merken

Nipy on functional brain MRI

Zitierlink des Filmsegments
Embed Code

Automatisierte Medienanalyse

Beta
Erkannte Entitäten
Sprachtranskript
welcome everybody who have a talk about by Unix on size you about brain imaging in base and these would be to be few thank you for thank you for coming on the morning on and moments in those of your time working the chemical because that Munich as the sole and
the outline of the talk is going to be the 1st thing went into
introduced the 9 titles for neuroimaging 1 going to give a brief short of a brief introduction to most of of the then I'm going to a vehicle of what this functional MRI and I'm going to explain a bit of a public database and and that's available to do this kind of experiments and I'm going to show the whole whole you can pre-process from right along with the learning
might so my body I'll is the is the is the name of a community is the common community of people that through neuroimaging and nearest science with Python and it's also module so there's also a module called my nite right so if you go to a knife I don't work you have done this in
a nice list of all the models available for neuroimaging so the first one I would say is and the nite model which allows you to open different types of neuroimaging file formats and so you have you have I O functions for the analyzed the gift gifting if T so these are the these are different types of neuroimaging files then you have a finite life which is a set of tools process of anatomical and functional MRI images certainly more than uh there's not 9 5 0 which is a which is a tool that allows you to define 5 points to process images it's
it provides a common interface to many other tools available spell
elsewhere so usually people to neuroimaging through command line or much love and 5 allows you to put
all these tools together it's very handy and then there's my time which has many
algorithms for time series analysis and the likewise which is to process of diffusion imaging so when you see these images of of hybrid tracts of the brain it's
with of diffusion so this this is a nice library of knife model to process this kind of thing and then you have new learning which is I think the new the
was a model which is done by the keynote speaker of this morning it it provides for us fast and easy
methods to do statistics with the with neuroimaging and it uses site learn so what this functional MRI functional MRI is uh uses the the bolts the goal the ball you know which is blood-oxygenation-level-dependent and it's and it takes in so it uses the difference in the magnetization of the blob flow when you have oxygen and not oxygen in in the in the blood so this allows you to
more-or-less measure where that teaching the brain is going on because neural neuronal neuron activity needs oxygen to work so this has some problems but it works and so the idea that you you require 1 whole volume in the interaction of this so the
idea is that you your career 1 whole volume each time you have usually you take every 2 seconds you take 1 home volume of the brain and the functional images is a time series or volumes of brain volumes and so there is an international neuroimaging data sharing initiative words which has many
interesting databases of with different sets of patients so you have patients
with some schizophrenia
alzheimer's and all
there they'd not all of them have a from right but most of them and you just you can download them and start
playing with this kind of of it of images so I'm going to show you a bit of a cold resonance data from my preprocessed so here I have 155 out this is the extinction of of this is effective so I use of and about to to load this is much we don't need a lot of good and low load function and this is this is the image data with some headers so but had there is necessary to to to put to put the brain in the correct coordinate system well here's here's a more most of them had the this is a long list and you have for example of the type of speech Fox the the 1st that the 1st 3 dimensions of the temporal limitation so this is 240 volumes all this find and and here so that each each voxel has 3 millimeters so it's 3 right the way through point state and so on and so this
image the image of the object of the get the of function which are stores the the this
volume from the from the images from the from the files so this this is unknown pirates and it has the shape that that we saw before and had a and so on 9 learning apart from this the physical parts of it has a very nice has some very nice supporting tools that the worker role with that I so import and in the plotting some model I use the index image function to to be the 1st volume of the of the resting resting image and to use this book and its function just to show you the 1st 4 the 1st volume of the functional image and I tried to plot the whole time series and the look but they couldn't do it with in not hopefully by was they heard that could use
of them OK but well I'm
using a tool that I II work I the work with and she this is a member of the functional image as you can see that there is a bit of movements there a bit so there's a bit of noise this image before Greenspan statistics with the they need to to correct for motion for the slice timing and his movies so I'm going to show a little bit of all of all this stuff so is less I mean
the motion correction is the 1st step on are produced this the about images that are usually you need to know what was the this slice
border of their position so for each volume of this kind of quiet but what about uh the the 1st 1 is less for a moment so the places have different kinds and since this is a lot of time related to data you need to you need to shift the slices in the correct order to make this this image look look at the and so this is dependent on the scanner so for example you can get for this type of the image is light
orders like this so it's alternating so 1st he doesn't know even some of their what are the odds of places and then the the ones and open and it alternates I'm sorry to to do this motion correction I'm using I and find it has the induced long from 1 to the other you have those space-time relying on the best that allows you to to do motion correction and slice timing correction so here here I create an instance of the subject of this class I I give the my function of the image I the repetition time which is also a scanning parameter so this means that the 2 seconds for each volume and his lifetime and is is the name of a function to to say what is this life or to acquire the talk where each slice of the soul of this class is going to search for the the function and applied to the image in order to to build those lights order of this
place on the list to perform the corrections so this and of In this case that's around dominant the I could use also w
to cachet these the result of this is a very nice tool for job with his is something very useful in neuro science to restore the steps the distance of their performance in the next and experiments so when I saw when use this process finishes get but a result which has the same shape of the work but now it's time to expect so I say that I use a knife I image through to and in this past and well I could walk let so 1st I'm going to I'm going to check the transformation that was performed so realign areas might be might the future class or that I used to to correct so the image and I think the terms both of formations and for its transformation I only only picking the translation part With transformation so it's transformation does so here I'm only using that I'm going to plot the translations so I have for each dimension explains that have those the translation that was applied to each volume of the of the rest of the image so this is the the final function defined matrix so the 1st transformation going of and let's take a current looks like not the following it's
not very easy to the difference is that if you take on the top top right there you see that there is much less movement than before you have to you have to well
with fMRI there are many paradigms in fact in the in the public datasets which
I showed you before and that you have a room you have task that from right where you you put the subject in the scanner and you ask you ask this this subject to perform different tasks and you get different
signals depending on the task and you can do statistics to see what areas are activating according to the to this tasks but also but also very low so this isn't
that another problem is resting state of from from right which should just leave the patient inside the scanner you ask them to not thinking anything and and you filter the image only measure low frequency signal in order to see how to bring he's working in rest so with this kind of measures you you can you can do some diagnoses itself so it's a very interesting modality to to work with I'm going to talk a bit about connectivity late so the 1st thing you have to do with this kind of thing with this kind of images you have to do about a band-pass filter you have to filter it with 3 with the
lower to their of your friends your very own critics and have frequency of the considerable I have this function here you know I think very ported from white cooperative from 5 which is something and not showing how to use here the copy the code of the functions to show you but it's too complex for so I applied this function to my to my realigned motion-corrected functional imagine I save it using the 2 final name of function which comes with the with the with me but all object this this function is returning the and object or object and if the image and they use that to find information to to say that is quite quite and and now let's look let's see how it looks now the so it's so
much messier but you see you see some of those what's the low frequency noise that monopolizing at t t i it's true that the I could mask is that of the brain areas to get only the voxels that from the brain and while this is so you can see use the following cannot see be something happening here but this is how call these images look like when in in the middle of the
process so of that 9 learned allows you to to calculate frame and it has this computes PPI mask to to calculate the brain mask from the from my from if it's and while it does quite
a good job if you have a clean a clean image you can also so I have here use of to just to check and Net and 90 and so the I learned and to to add to the humor of the brain mask as edge here in the over the 1st volume of the resonant peak in so and another thing that we can do is to use the anatomical image we usually when you do functional MRI you also acquire an
anatomical image is also called the 1 way to and period and if
it's there in the middle MRI modality that gives you the best resolution and tissue contrast so it allows you to do anatomical scan segmentation from the brain and
you can use that information over the functional image so for example and shown here are some of an anatomical
image from the core problem with that and I would have to register this image to the functional image through the the to work with segmentation but they do here in the functional alignment of that there some on registration algorithms listen to to by my tools but I use all the all the tools from
all the running much another step important step is is moving since the these images are only the the department to do was moved move them and so on I learned that gives you very nice
functions it's called was moving image and you can so this you give me the file so this is the motion-corrected 2 thirds of the image and you give them the full width half maximal so is the size of the of the kernel to to build small things In the limit and no let's see how the
world is moving is moving image looks like so it's like
and what sometimes happens I'm going to change the contrast and now it looks a little bit better and you can see the the French sought to the rest of the world the resting state part of the functional image so
connect each analysis so connectivity analysis is that when you take this filtered months of functionally image and you select different regions usually under different anatomical images you extract the signal of each of these regions and you calculate the covariance between this this regions so what variants of the resting state between these regions and for that you usually need brain atlas of Annapolis gives you out
regions in the brain the anatomical or functional but relevant so I learned in the effects of someone who has for example the fact atlas Forward Bloc function which allows you to to get a nice and nice apple down the showing the regions of these of and you can use each of these regions to to perform collective challenges so on my
learning is a very nice it's very well documented tools so yeah so but it
uses sighted Learning and who performed the covariance and measure and here she is
users connectivity the matrix source in each in each pixel of the image you have worked on covariance measure between the different regions every other region of the of the brain of the list of the children and the pulsar also between to plot the from the connections between the regions so this is very used to to to compare different types of
of patients a subjects in different so there's another very interesting thing about arresting state is that it was it has been discovered that 1 of the brain during rest and it has connected regions that work when you arrested so these are the the most
the most famous ones people are still doing research on over this this kind of this kind of knowledge and so these are the different areas of the brain that are known to work together when you are resting and you can use this for example as well
to who check how well it works in patients with Alzheimer's and other neurodegenerative diseases with learned the composition you can perform ICA which is
independent component analysis and they have so that it's called canonical and also the learning and you can use this to extract of these resonant peak of components of the functional image the you have to do the
work works so for a small how I do it import from middle name and the composition of the learning I have to to set the number of components I want to extract from the from the rest in of the image so basically this all of so these are the most significant components of the of the of the whole
volume of the whole rest since the volume and you can also to apply this to your data and you get the components so here there is a very nice example from I learned so this is once they expect to use this
components a lot of over a template and these are the areas where this component is present for you get different of all these different resting-state network so well I'm using
my point of of the time I'm a so of model it's called white which has many many of these phone calls literature showed you very easy to use with all of the dataset so you can continue pipes to run over they and it has it has workflows 2 different types
of images and so yeah that's that's all you have any questions you need any help with this kind of analysis these come
to me any time what thank
you thank solution to any questions thank you for talk I have a couple of questions from new and he show that the mean of the turks some Margaret install small thing making sense to me that I wanted to make this something specific about brain images that for instance can ecological psyche the nature of the and more generate the image processing by race and the other 1 is that have the that mark the used as is reading used toward you in each position and I would like to know how this all these tools compared to commercial temperatures grateful there there is some of the specifics of this type the images regardless of what the law calls for example brain the lingering image 2nd image sort of so for example for diffusion image you need very specific models and feed to your data for functional image you could so psychic the image is not really prepared for 40 I mean you can do it but for example you're working with China for so for example to do was moving things if you run out of small thing with the image so for example as is missing you need to know the voxel size because when working in mm for example now that there's some specificities that the image is not it's not wouldn't be of practical and so yeah not all of these values that there they're all the tools that the command line varies from different universities the thing is that if you want to work with this I mean this European research with and you need to publish so and and so if you want to publish you have to use a tool everybody knows otherwise they're going you're going to get a lot of critics from the reviewers yeah so for example there's a very nice tool from University College London which is called the physical parameters mopping which were totally on much of it has a gooey so it's easy to use of the good thing is that for example my wife has an interface for most of the useful things that at offers and the now that those that that many is another world a lot of tools available to do this most of them are probably further questions I I was wondering how loyal they may become research and software development and and doctor I'm going to make development goal kind of dynamin call ghost 1st and what's the relation of the I mean I guess brought the friend of course different professionally byzantine but and how do they interact with 2 right there is this useful where how everyone can be doing something and push it to the other how does it work it doesn't work it does it work out what the so yeah it's a multidisciplinary disciplinary fields where you have to do this is mathematicians have colleges have medical doctors if you have computer centers of half of so the depends on the application for which they did you want to do during the lab where they do all psychology that psychology experiments mostly the 2nd part that could be of use allergies who is leading the ideas or psychology medical doctor who could check whether your your data is saying something to them because they're the ones that of all of the diseases of brain of course in the end I have on my computer scientists and have to learn the object of brain anatomy of understanding of the of of but most I would say computer scientists are going on the of thank you for talk and actually I was wondering what and these particular fields you need so many tools shut off active in the beginning of talk you showed us different tools what but all of them are imporant and that's gonna just for a semantic so we have so in the MR scanner is very flexible the can see you can configure it to run different types of measures from the brain and each of these measures they are wonderful view of of research on its own so you have diffusion you have functional you have chemical imaging so all these fields they are so all this modality so the era they have different types of another process I think stage octave difficult to talk about computer science in the US and it's in 1 slot on and wondering if you guys have any training and if there is a chain if you guys have developed a new training specifically for scientists to get involved in this rather than computer scientists who were trying to learn science so to speak float I have been well I tried I mean maybe to I'm doing software carpentry and they included and 95 and by tutorial 1 thing to have way life and people enjoyed it so because mostly because they know the people that are on the eve so of the as a matter of tools the common line tools but I wanted to show them how to use in my life because the useful for now that's it I mean I'm not paid to run out of the carpentry or In fact I have get on holidays yeah something like this you don't know more Christians I have 1 that you mentioned that that's less off connected your brain activity and what does this what to do D entries and this protest represent exactly is it a special action that the body to room but for them you so there is a net loss of brain gain brain activity here if we take an entry is at last like is it is it would be if I understand well uh some regions of the brain which would be active at the same moment yeah so it's yeah so it's how all of the different areas work together when when the so for example mean connectivity in the resting state connectivity you're measuring how synchronized region OK and then what does that have presented can you ensure that 2 bodies doing special movement or no it's resting so the subject is part in this kind of thing look you can you can see for example whole hollowing Alzheimer's Disease patients of some areas that are not there is going to on the connection is different or lower the following OK thank you thank you and Questions from that as well Our having had your friends can manage that if did it look familiar looks like this this yeah so I you have to be here so that is 1 of my brain scans this this is merely
inside and this is the of hash thank you very much and
without
Metropolitan area network
Font
Momentenproblem
Computeranimation
Lineares Funktional
Bit
Modul <Datentyp>
Rechter Winkel
Modul
Computeranimation
Schnittstelle
Lineares Funktional
Videospiel
Prozess <Physik>
Punkt
Mailing-Liste
Extrempunkt
Elektronische Publikation
Computeranimation
Metropolitan area network
Informationsmodellierung
Menge
Datentyp
Dateiformat
Gravitationsgesetz
Schnittstelle
Metropolitan area network
Schnittstelle
Prozess <Physik>
Algorithmus
Zeitreihenanalyse
Hybridrechner
Analysis
Computeranimation
Metropolitan area network
Lineares Funktional
Web Site
Subtraktion
Statistik
Informationsmodellierung
Programmbibliothek
Gravitationsgesetz
Datenfluss
Computeranimation
Meta-Tag
Zeitreihenanalyse
Zwei
Interaktives Fernsehen
Übergang
Wort <Informatik>
Spezifisches Volumen
Einflussgröße
Computeranimation
Gammafunktion
Dualitätstheorie
Subtraktion
Bit
Resonanz
Punkt
Hausdorff-Dimension
Sprachsynthese
Computeranimation
Metropolitan area network
Datentyp
Meter
Inverser Limes
Spezifisches Volumen
Gravitationsgesetz
E-Mail
Große Vereinheitlichung
Bildgebendes Verfahren
Gammafunktion
Lineares Funktional
Datenhaltung
Güte der Anpassung
Mailing-Liste
Physikalisches System
VOXEL
Menge
Last
Rechter Winkel
Datenerfassung
Aggregatzustand
Objekt <Kategorie>
Metropolitan area network
Lineares Funktional
Informationsmodellierung
Zeitreihenanalyse
Automatische Indexierung
Mereologie
Softwarepiraterie
Spezifisches Volumen
Elektronische Publikation
Speicher <Informatik>
Bildgebendes Verfahren
Computeranimation
Bit
Statistik
Program Slicing
Geräusch
Vorlesung/Konferenz
Bildgebendes Verfahren
Momentenproblem
Datentyp
Program Slicing
Multitasking
Spezifisches Volumen
Ordnung <Mathematik>
Bildgebendes Verfahren
Computeranimation
Lineares Funktional
Parametersystem
Videospiel
Unterring
Program Slicing
Klasse <Mathematik>
Zwei
Vorzeichen <Mathematik>
Mailing-Liste
Oval
Aliasing
Computeranimation
Metropolitan area network
Offene Menge
Spezifisches Volumen
Gravitationsgesetz
Ordnung <Mathematik>
Bildgebendes Verfahren
Brennen <Datenverarbeitung>
Instantiierung
Resultante
Matrizenrechnung
Lineares Funktional
Shape <Informatik>
Prozess <Physik>
Hausdorff-Dimension
Vorzeichen <Mathematik>
Transformation <Mathematik>
Term
Computeranimation
Flächeninhalt
Prozess <Informatik>
Mereologie
Translation <Mathematik>
Spezifisches Volumen
Abstand
Gravitationsgesetz
Bildgebendes Verfahren
Subtraktion
Rechter Winkel
Programmierparadigma
Computeranimation
Task
Statistik
Task
Flächeninhalt
Rechter Winkel
Aggregatzustand
Computeranimation
Einfach zusammenhängender Raum
Lineares Funktional
Bit
Frequenz
Code
Computeranimation
Metropolitan area network
Modallogik
Rechter Winkel
Information
Ordnung <Mathematik>
Einflussgröße
Bildgebendes Verfahren
Aggregatzustand
Flächeninhalt
Geräusch
VOXEL
Verdeckungsrechnung
Spezialrechner
Metropolitan area network
Rahmenproblem
Prozess <Informatik>
Bildauflösung
Spezifisches Volumen
Bildgebendes Verfahren
Computeranimation
Verdeckungsrechnung
Bildauflösung
Extrempunkt
Computeranimation
Modallogik
Spezialrechner
Metropolitan area network
Algorithmus
Registrierung <Bildverarbeitung>
Speicherabzug
Kontrast <Statistik>
Information
Bildgebendes Verfahren
Große Vereinheitlichung
Bildauflösung
Metropolitan area network
Lineares Funktional
Cliquenweite
Inverser Limes
Elektronische Publikation
Bildgebendes Verfahren
Computeranimation
Kernel <Informatik>
Kovarianzfunktion
Bit
Mereologie
Vorlesung/Konferenz
Kontrast <Statistik>
Dialekt
Bildgebendes Verfahren
Computeranimation
Analysis
Aggregatzustand
Soundverarbeitung
Metropolitan area network
Lineares Funktional
Extrempunkt
Gravitationsgesetz
Große Vereinheitlichung
Dialekt
Computeranimation
Einfach zusammenhängender Raum
Matrizenrechnung
Kovarianzfunktion
Pixel
Inverse
Singularität <Mathematik>
Mailing-Liste
Quellcode
Dialekt
Computeranimation
Vorhersagbarkeit
Metropolitan area network
Einflussgröße
Bildgebendes Verfahren
Große Vereinheitlichung
Gammafunktion
Flächeninhalt
Dialekt
Computeranimation
Eins
Aggregatzustand
Metropolitan area network
Karhunen-Loève-Transformation
Resonanz
Stochastische Abhängigkeit
Zusammenhängender Graph
Bildgebendes Verfahren
Computeranimation
Metropolitan area network
Diskrete-Elemente-Methode
Zahlenbereich
Zusammenhängender Graph
Extrempunkt
Bildgebendes Verfahren
Computeranimation
Gammafunktion
Metropolitan area network
Datennetz
Flächeninhalt
Template
Zusammenhängender Graph
Spezifisches Volumen
Computeranimation
Informationsmodellierung
Typentheorie
Datentyp
Hilfesystem
Computeranimation
Analysis
Subtraktion
Einfügungsdämpfung
Prozess <Physik>
Wellenpaket
Gewicht <Mathematik>
Momentenproblem
Ortsoperator
Natürliche Zahl
Gruppenoperation
Kartesische Koordinaten
Computer
Gesetz <Physik>
Computeranimation
Informationsmodellierung
Software
Konstante
Datentyp
Installation <Informatik>
Softwareentwickler
Informatik
Grundraum
Gerade
Einflussgröße
Bildgebendes Verfahren
Schnittstelle
Einfach zusammenhängender Raum
Umwandlungsenthalpie
Videospiel
Parametersystem
Lineares Funktional
Sichtenkonzept
Relativitätstheorie
Systemaufruf
Bildanalyse
VOXEL
Quick-Sort
Dialekt
Arithmetisches Mittel
Objekt <Kategorie>
Modallogik
Verkettung <Informatik>
Datenfeld
Flächeninhalt
Mathematikerin
Mereologie
Schwimmkörper
Oktave <Mathematik>
Aggregatzustand
Instantiierung
Metropolitan area network
Hash-Algorithmus
Computeranimation

Metadaten

Formale Metadaten

Titel Nipy on functional brain MRI
Serientitel EuroPython 2016
Teil 158
Anzahl der Teile 169
Autor Savio, Alexandre
Lizenz CC-Namensnennung - keine kommerzielle Nutzung - Weitergabe unter gleichen Bedingungen 3.0 Unported:
Sie dürfen das Werk bzw. den Inhalt zu jedem legalen und nicht-kommerziellen Zweck nutzen, verändern und in unveränderter oder veränderter Form vervielfältigen, verbreiten und öffentlich zugänglich machen, sofern Sie den Namen des Autors/Rechteinhabers in der von ihm festgelegten Weise nennen und das Werk bzw. diesen Inhalt auch in veränderter Form nur unter den Bedingungen dieser Lizenz weitergeben
DOI 10.5446/21090
Herausgeber EuroPython
Erscheinungsjahr 2016
Sprache Englisch

Inhaltliche Metadaten

Fachgebiet Informatik
Abstract Alexandre Savio - Nipy on functional brain MRI This is an introductory talk to modern brain image analysis tools. I will show how to use nipy tools to process one resting-state fMRI subject, perform intra-subject registration, ICA analysis to extract and visualize resting-state networks. If the time allows me I will introduce how to perform non-linear registration to to atlas-based segmentation. The outline of the talk: 1. Present the COBRE dataset and show its characteristics. 2. Use nibabel to open a NifTI file and see the matrix/volume parameters. 3. Use nilearn.plotting to show the anatomical image. 4. Use nipy to co-register the anatomical image to the fMRI image. 5. Use nilearn to perform CanICA and plot ICA spatial segmentations. If time allows: 7. Present a brain anatomical atlas and its template. 8. Present the tools needed for non-linear registration. 9. Show the result of an atlas-based segmentation result. 10. Use nilearn to calculate the resting-state functional connectivity matrix of the subject. 11. Plot it with Bokeh.

Ähnliche Filme

Loading...