For nearly two decades resting-state fMRI has ruled supreme. And resting-state is suprisingly powerful. But we can do so much better by using broad batteries of tasks with the right design. Here a double-punch delivered by two papers from @carobellum.bsky.social and @basselarafat.bsky.social!
Our companion paper led by @basselarafat.bsky.social is now out as a preprint, introducing a toolbox to optimally select tasks and run them in the MRI scanner: doi.org/10.64898/202...
Are you thinking about acquiring multi-task fMRI for your study but don’t know where to start?
@bassel_arafat
just published a toolbox to optimally select tasks and run them in the MRI scanner - with many commonly used tasks ready to go! Check out the preprint below 👇
This study looks great - this issue has finally been getting the attention it deserves - if network architecture is ‘intrinsic’ it should say something about connectivity across tasks …
Thank you Colton!
Really great work demonstrating the importance of task-based responses in characterizing the functional organization of the brain. Congrats to @carobellum.bsky.social and the rest of the team!
👇
What would you consider the standard? In human neuroimaging my read of the literature is the field leans quite heavily on resting-state.
We also thought implementation and customization would be a big bottleneck, so we have been thinking about ways to help streamline task choice and implementation for a given clinical or research goal. There is a companion paper led by our talented PhD student Bassel Arafat coming out very soon...
Thank you! Agreed, that would be super interesting - perhaps something we can try in a follow up.
@carobellum.bsky.social really wanted to know how well resting-state correlations can predict which areas co-activate over a wide array of tasks. It turns out that you are better off running a short Multi-task battery, even if the tasks are completely different from the ones you want to predict!
Cool demonstration that using multiple tasks (diversity is important) may be more informative than resting state for uncovering individual functional connectivity patterns
Diversity matters. Multi-task, multisensory, crossmodal.
Thank you!
Task fMRI for the win! This is such great work, and a brilliant explainer thread too.
100% agree. Until we can train infants to hold a button box, rest will never be superseded for some populations :) But I would think multi-task is definitely worth considering for standard scanning protocols in healthy adult populations!
Yes, I love that paper from Jingnan and Randy, they show super neat how rest and task can be similar to each other.
With limited data or acquisition time, though, the multi-task approach seems really powerful to get you the most bang for your buck (and sadly scanning time still requires many bucks)
I love this idea! Have to give it a read
"Resting-state can be highly reliable. But reliability does not guarantee validity."
So my interpretation of the results is less that rest gets at a state that is fundamentally different from task, but that rest is one state of many. And it's great to reveal *some* organization, but it just doesn't generalize to many other states, which we actively go through in daily life.
I would also say there are tasks that we included that will get close to what people might 'naturally' do in the scanner and the complexity of mind-wandering: Thinking about their childhood home (that's literally a task), thinking about other people, watching a movie, generating words, remembering..
We took the task time series and just treated it as rest, i.e. we threw all the info on what people were doing away and then compared to rest. Still, task won! Which means even if you completely ignore your task info, you reveal the organization better with task because you are driving the system.
Great question, and that's exactly what we wondered! Is it that we don't know what people are doing (maybe considered "noise", or just some missing info), or is it actually that whatever they are doing, we are "driving" them more into like in task? So we did a careful comparison to tease this apart:
Thanks so much! We haven't, but it would indeed be a great extension...
This looks awesome and I will now be having all subjects play WarioWare in the scanner
Would love to be a volunteer!
Thank you so much!
Diversity of tasks>>resting state for predictive power.
Check out our new preprint!
We directly tested a common assumption in neuroimaging:
Is resting-state the best estimate of intrinsic organization?
Spoiler: Across datasets and brain systems, multi-task fMRI did better.
👇