Goals Get a handle on how fMRI can be used to ask interesting questions about the brain Learn to play music Get some practical fluency in how to do fMRI analysis, including pattern-based analyses Learn to play scales Become a better-informed reader of fMRI papers and news reports Not every #1 hit is the Beatles. But some are

My favourite science quote It’s better to have an approximate answer to the right question than an exact answer to the wrong question - John Tukey

What’s a good question?

What’s a mechanism? Eye movement demo

What’s a mechanism? (continued) From outside the head, we can figure out something about what’s going on inside the head So, if we look inside the head, we must have a royal road to finding mechanisms, right?

Problem: This picture does not show any neural mechanisms 6

Overview of course Main goal: course project Analyse a publicly available dataset Design a new expt Or something else that we agree on 40% of grade, but there’ll be sub-chunks Might even turn into a publishable paper! Computational exercises (20%) Mini-essays (20%) Tweet-length Qs about readings (10%)

Thanks for the feedback from mini-Qs Helped me to see a couple of frequent stumbling blocks Blood flow vs Blood oxygenation Reverse inference Please e.mail as plain body-text if possible (not Word attachments). Not a big deal, though Anybody send in answers that I didn’t reply to yet?

Piazza My hope: good for class discussion Feel more than free to post anonymously I’ll be posting polls, seeking feedback Quick Q: who here in room isn’t on my mailing list? Who didn’t get a Piazza invite Will use to figure out good office hours time

Computer practicals Hands-on experience doing data analysis Will be using Matlab and SPM Can also use your own favourite tools if you like With the invaluable assistance of Dan Cole Let’s try all to squeeze into one session

MRI Magnetic Resonance Imaging Takes a 3D picture of the inside of body, completely non-invasively One picture, just shows the structure

fMRI functional Magnetic Resonance Imaging Shows brain activity (indirectly) Takes a series of pictures over time, e.g. one every three seconds The “f” in fMRI means functional, i.e. you get a movie of brain function, not a still image of brain structure

What are we actually measuring with fMRI? An MRI machine is just a big magnet (30,000 times stronger than Earth’s magnetic field) The only things it can measure are changes in the magnetic properties of things inside the magnet: in this case, your head When neurons are active, they make electrical activity, which in turns creates tiny magnetic fields BUT far too small for MRI to measure (100 million times smaller than Earth’s magnetic field) So, how can we measure neural activity with MRI?

What makes fMRI possible: Don’t measure neurons, measure blood Two lucky facts make fMRI possible When neurons in a brain area become active, extra oxygen-containing blood gets pumped to that area. Active cells need oxygen. Oxygenated blood has different magnetic properties than de-oxygenated blood. Oxygenated blood gives a bigger MRI signal End result: neurons fire => MRI signal goes up This fMRI method is known as BOLD imaging: Blood-Oxygenation Level Dependent imaging. Invented in 1992.

But neurons do the real work, not blood. Neurons represent and process information Individual nerve cells (neurons) represent information Sensitive to “preferred stimuli”, e.g. /ba/ These stimuli make them active Firing activity: send electrical spikes to other neurons /ba//ba/-sensitive neuron

Populations of neurons process information together Information is distributed across large populations of neurons, and across brain areas There’s no “grandmother cell”: the one single cell that recognizes your grandmother To really understand the brain, we’d need somehow to read the information from millions of individual neurons at once!

The basic design of an fMRI experiment Aim: Find which brain areas are active during a given task E.g. discriminating speech sounds, producing speech Typical design: Present blocks, e.g. 30s of task, 30s of rest Measure fMRI activity regularly every few seconds Look for brain areas which are more active during the task periods, compared to rest periods

Example time-courses MRI signal from voxel that correlates well with task: Active Task TIME Time-course of task versus rest periods Rest Task Rest Signal from voxel that does NOT correlate with task: Inactive

What are those little coloured blobs, actually? Colour represents statistical significance of how well the voxel’s activation correlates with the task. The hi-res grayscale anatomical picture underneath the coloured blobs is a completely different type of image, from a different type of scan. Shows the anatomy at the spot where the significant voxel’s time- course was recorded.

The key problem Interpreting what brain activation means

Reverse inference Why it’s hard to infer processing from activation: Brain areas are multi-functional Attention Intention Spatial reasoning Numerical magnitude Parietal cortex ???

A famously horrible example “You love your iPhone, literally” “But most striking of all was the flurry of activation in the insular cortex of the brain, which is associated with feelings of love and compassion. The subjects’ brains responded to the sound of their phones as they would respond to the presence or proximity of a girlfriend, boyfriend or family member. In short, the subjects didn’t demonstrate the classic brain-based signs of addiction. Instead, they loved their iPhones.”

Distinct representations may produce same overall activation Speech area /ba/-sensitive population of neurons /da/-sensitive population of neurons Subtraction: /ba/ minus /da/ = zero

Representations that standard fMRI can handle: Localised and segregated If you do a PubMed search for representations and fMRI, you find figures like these: Hand Lips Faces Houses

Representations that are more difficult: Distributed and overlapping

Distinct but overlapping representations: same average activation, but different local patterns Average local activation is the same for both stimuli Stimuli A and B activate the same neural population, both activating shared neurons to differing degrees, so they elicit different activation patterns Stimulus AStimulus B

Representational pile-up Standard analysis: Smoothed local average activations all look the same Pattern-based analysis Maybe can tell the unsmoothed spatial patterns apart