Beginner’s guide to

computational neuroscience

Step 1. If you are a beginner to programming for research I highly recommend checking out the Good Research Code handbook. This comprehensive resource will walk you through setting up your computer and project workspace with suggestions that may save you hours of headache down the track (something I wish I had when I started in this field!).

Study design considerations: You will cover some of these things in Step 2, but here are some resources that may help when it comes to data management and statistical analysis.

Step 2. If you are new to computational neuroscience and/or magnetic resonance imaging (MRI) and would like to wrap your head around the concepts, see the below resources and authors to fast track your learning. If you would like to jump right into neuroimaging analysis head to Step 3.

  • Neuromatch Academy - this free, extensive resource provides an introduction to concepts and machine learning for computational neuroscience. Their youtube videos offer maths refreshers (i.e. in algebra, geometry) and the specific knowledge of dynamic and stochastic processes which you will need to consider in your analysis. You can adapt this knowledge to studies of functional (EEG, fMRI) and structural (diffusion, volumetric MRI) data which is why I highly recommend this course. They also hold paid courses where you can work in a team and get expert help, but all the content is also freely available online (again, wish I had this when I started out!).

  • “Modeling Neural Circuits Made Simple" by Robert Rosenbaum - based on the book by the same name - an introduction to modeling neural data from the single neuron level (static and timeseries) with code and explanations.

  • Introduction to Neuroimaging Analysis - a free “primer” series that takes you through from start to finish, beginning with image segmentation, tissue types, different analyses (DTI, TBSS - there are newer techniques but traditional ones are also important to know about), brain atlas registration and so on.

  • MRI from Picture to Proton - an invaluable introduction to the history of and different types of MRI.

  • After viewing the above resources, my suggestion is to first understand whether you have structural or functional data, and at what scale (micro to macro - this site by the Human Brain Project has a good explanation of the various scales at which computational neuroscience and brain simulation activities can be conducted/examined). Next, think about the type of analyses you will do i.e. if you only have functional data (fMRI, EEG) - will you do time series or causal analyses? If you have both data types - do you plan to use them individually/together? Incorporate these type of questions into your study design planning.

  • Some (not all) key authors/resources (in no particular order) for an expert introduction into network neuroscience:

Step 3. Neuroimaging analysis for structural (i.e. diffusion MRI, volumetric MRI) and functional (EEG) imaging. This is a giant field and it is easy to get lost / waste time using tools that may not do the entire pipeline, which you only find out after hours setting them up.

In order to select the right tools, data and study design to produce the outcome you are after, you may wish to ask yourself a few key questions before you start:

  • what is your data acquisition protocol? i.e. what is the MRI sequence you used to obtain the data? the acquisition protocol will determine the best tools and analysis i.e. if you have single shell diffusion data with 32 directions you can use MRTRIX for most analyses but you will not be able to do analyses like tractometry or differential tractography in DSI Studio as it requires multishell data.

  • what type of analysis do you want to do? tractography? connectomics?

  • how many subjects do you have in your a. disease group (assuming you have one) b. healthy controls etc

The below resources may help you answer these questions and the toolboxes/software come with extensive documentation to guide you. I‘ve categorised the programs into the data types but you will need to do a little digging to determine which are suitable for the analysis you want to do and the data/time/resources and programming skillset you have.

Structural and functional data:

Data processing and analysis tools

  1. Suggest starting with Neurodesk - a fabulous new tool which will save you hours of time downloading the multiple softwares you will need for each step of neuroimaging analysis. Neuroimaging analysis requires downloading a number of software (Freesurfer, AFNI, ANTs) to process the raw data before you can even start your statistical analysis. Neurodesk now enables you to skip this step, runs on any operating system (Mac, Windows, Linux) and potentially even work directly from your browser.

  2. MNE - Open-source Python package for exploring, visualizing, and analyzing human neurophysiological data: MEG, EEG, sEEG, ECoG, NIRS, and more.

  3. Brainstorm - Brainstorm is a collaborative, open-source application dedicated to the analysis of brain recordings: MEG, EEG, fNIRS, ECoG, depth electrodes and multiunit electrophysiology.

  4. The Virtual Brain (TVB)- Possibly the most comprehensive tool for analysing all types of structural and functional neuro data at all scales. You cannot do pre-processing in this tool, but once you have processed data, you can do everything from analysis to visualisation in this tool. It takes some time to set up and learn but the website contains all the tutorials you need to get started.

  5. Nipype - similar concept to Neurodesk but slightly older. Adding this here as it is an option but it is not clear if this toolbox is regularly maintained so you may wish to stick to Neurodesk.

Data analysis only (i.e. when you have pre-processed data already)

  • BATMAN tutorial (Basic And Advanced Tractography with MRtrix for All Neurophiles) - My number 1 recommendation for introduction to diffusion neuroimaging analysis - this is the only tutorial you need to follow to learn tractography and analysis and visualisaion of results using the MRTRIX software (linked below). If you want to learn about the latest approaches in diffusion weighted imaging (DWI), this is where you want to be.

  • Brain Connectivity Toolbox (BCT) - graph theory analyses using connectomes (any data type).

  • Network based statistics - Popular Matlab toolbox for testing hypotheses about the human connectome.

  • NBS Predict - an extension of NBS but with several options for machine learning (any data type in matrix format).

  • Dyconmap - a neuroimaging python module specifically designed for estimating the dynamic connectivity and analyzing complex brain networks; from neurophysiological data such as electroencephalogram (EEG), magnetoencephalography (MEG) and functional magnetic resonance imaging (fMRI) recordings. It includes several submodules to work with, such as chronnectomics and graph-theoretical algorithms, (symbolic) time series and statistical methods. Paper here .

  • Graphvar - “GraphVar” is a user-friendly graphical-user-interface based toolbox (MATLAB) for comprehensive graph-theoretical analyses of brain connectivity, including network construction and characterization, statistical analysis (GLM and Machine Learning) on network topological measures, and interactive exploration of results. It combines features across multiple current toolboxes, such as the Brain Connectivity Toolbox, Network Based Statistic Toolbox, BRAPH, and BrainNetClass and represents a comprehensive collection of graph analysis routines for functional neuroimaging researchers.

Structural data: Tools and resources

  • MRTRIX3 - MRtrix3 provides a set of tools to perform various types of diffusion MRI analyses, from various forms of tractography through to next-generation group-level analyses. It is designed with consistency, performance, and stability in mind, and is freely available under an open-source license.

  • DSI Studio - DSI Studio is a tractography software tool that maps brain connections and correlates findings with neuropsychological disorders. It is a collective implementation of several diffusion MRI methods, including diffusion tensor imaging (DTI), generalized q-sampling imaging (GQI), q-space diffeomorphic reconstruction (QSDR), diffusion MRI connectometry, and generalized deterministic fiber tracking. NOTE: you will need certain data acquisitions to do analyses with this toolbox so check first what is required.

  • Computational Anatomy toolbox - This toolbox is a an extension to SPM12 (Wellcome Department of Cognitive Neurology) to provide computational anatomy. This covers diverse morphometric methods such as voxel-based morphometry (VBM), surface-based morphometry (SBM), deformation-based morphometry (DBM), and region- or label-based morphometry (RBM).

  • Handbook of Structural Brain MRI analysis by Jerome Maller - VERY handy and easy to understand introduction to structural MRI analysis.

  • Brainspace - when you discover that connectome analysis draws distinct boundaries between brain regions (“nodes”) but in real life, the functional brain activity does not consider such physical boundaries, you may wish to use this toolbox to analyse and convery results on a “gradient” which removes the need for hard boundaries.

  • Radiomics - Radiomics is a newer area and mostly used where lesion data is concerned. From this paper: “Radiomics aims to quantify phenotypic characteristics on medical imaging through the use of automated algorithms. Radiomic artificial intelligence (AI) technology, either based on engineered hard-coded algorithms or deep learning methods, can be used to develop noninvasive imaging-based biomarkers.” This toolbox is an open-source python package for the extraction of Radiomics features from medical imaging.

  • Connectomemapper - Connectome Mapper 3 is an open-source Python3 image processing pipeline software, with a Graphical User Interface, that implements full anatomical, diffusion, resting-state MRI, and EEG processing pipelines, from raw Diffusion / T1 / BOLD / preprocessed EEG to multi-resolution connection matrices , based on a new version of the Lausanne parcellation atlas.

NOTE: if you are doing diffusion MRI analysis as part of a research lab you may wish to follow the ISMRM’s “best practices - diffusion MRI”. I wish I could share them here but they are only available to members. Check with your supervisor / boss about how to join ISMRM and then search those terms.

Functional data: Tools and resources

Open source data:

You will no doubt come across many open access datasets however below are some large, comprehensive datasets of structural and functional data that you can get started with easily.

NOTE: If you are new to coding/programming or even if you have a beginner level knowledge of python / R, check out the below resources:

  • W3 Python tutorials - these are probably the best I’ve come across for beginner level coders needing assistance

  • R - the main website for all R programming tools

  • Braingraph

  • Pydicom

Other important resources:

Machine / deep / federated learning resources

If you need more beyond the Neuromatch Academy info (Step 2), there are many free courses i.e. on Udacity, Udemy, and the actual sites of key popular programs such as Tensorflow / Pytorch, which you can search online. With a little help from ChatGPT 3 / 4, you will likely learn machine learning basics in no time at all. The below resources include some tools which I used as a beginner and some which facilitate a move towards the future through federated learning in research and decentralised science.

  • Google crash course in machine learning

  • Deep learning onramp - MATLAB based starter course

  • Courses by Mike Cohen - if you can pay, he offers incredibly beginner friendly courses on the necessary maths, and programming skills needed for neuroimaging analysis. A ton of free lectures are here.

  • Gradiohub - easy to learn open source machine tools

  • Openmind - courses and community for learning open data science and federated learning

  • COINSTAC - Collaborative Informatics and Neuroimaging Suite Toolkit for Anonymous Computation (COINSTAC). A federated, decentralised pipeline as described on their site: “COINSTAC is software to foster collaborative research, removing large barriers to traditional data-centric collaboration approaches. It enables groups of users to run common analyses on their own machines over their own datasets with ease. The results of these analyses are synchronised to the cloud, and undergo aggregate analyses processes using all contributor data.”

Data visualisation

You will want to show off your cool analyses in nice graphs and images that are simple and make sense but look great. I suggest having a specific tool/program or several where you have code which can be reused - almost all papers in computational neuroscience will use some / all of the following graph styles (and sometimes a blend).

  • heatmap

  • boxplots/violin plots

  • histogram

  • error bars and error bars with before and after trends

  • scatterplots

  • density plots

  • dendrogram

  • chord plot and variations

My data visualisation suggestions if you are a student who plans on writing academic papers:

See below for a list of free toolboxes/programs/resources that I found useful. If you are programming pro (in python or R) or you can get what you need from the above tools (MRTRIX, DSI Studio, TVB, etc) you may not need these.

  • Brainnetviewer

  • Pysurfer - PySurfer is a Python library for visualising cortical surface representations of neuroimaging data.

  • Dataviz project - helps you workout what graph you need based on your data.

  • Data to viz - as above, helps guide you to the correct visualisation for your data.

  • Gephi - Gephi is an open-source software for visualising and analysing large networks graphs. Gephi uses a 3D render engine to display graphs in real-time and speed up the exploration. You can use it to explore, analyse, spatialise, filter, manipulate and export all types of graphs.

  • Bokeh - Bokeh is a Python library for creating interactive visualisations for modern web browsers. It helps you build beautiful graphics, ranging from simple plots to complex dashboards with streaming datasets. With Bokeh, you can create JavaScript-powered visualisations without writing any JavaScript yourself.

  • Python graph gallery - chord diagrams are often used for connectome representation - this is a handy resource to create them.

  • GGPlot - ggplot2 is an R package for producing statistical, or data, graphics.

  • Brainrender - an older python based software for visualisation of neuroanatomical and morphological data.

  • Flourish - low/no code data visualisation - simply upload your data spreadsheet. Lots of templates!

  • Streamlit - Streamlit turns data scripts into shareable web apps in minutes. All in pure Python. No front‑end experience required.

  • Plotly - a paid resource and good for making dashboards (there is some free but limited functionality).