Man has himself MRIed twice a week for a year and a half, plus tracking a lot about his life. The data mining is still going on, but at least it’s been shown that (probably) people’s connectomes change pretty rapidly.
I’m also posting this to the media thread because I’m not sure where it’s more likely to be seen.
Calling what was mapped here a ‘connectome’ is REALLY stretching it. When they make those graphs of parcels connected to each other, what they’re doing is just measuring the correlation between activity as revealed by an fMRI (which is itself removed from activity, measuring the short-term fluctuations in bloodflow as a result of energy requirements) in different parcels of brain and drawing a ‘connection’ when the coefficient is high enough. Correlation is not just connection.
I do note that there was diffusion tensor imaging (which shows you the average orientation of fibers in any given voxel [and showed an unusual crossing mixed fiber feature in a spot of his corpus callosum and will probably show similar oddities throughout the brain in any given human] ) and I will try to get at that information once I am past a paywall later on, but the repeated MRIs appear to be fMRIs.
MRIs: a lot of different scans including 15 T1 and T2 weighted structural scans; 19 diffusion-weighted scans. fMRI was mostly resting state (100), but also included various tasks such as n-back (15x), motion discrimination/stop signal (8x), object localiser (8x), verbal working memory localiser (5x), spatial WM (4x), breath holding (18x).
Oops! In whatever case I’m back from my situation-with-effectively-only-mobile-internet-access and have the paper now.
The repeated scans were indeed fMRIs measuring correlation of metabolic activity (a good proxy for activity) under various conditions. They made one diffusion tensor map from all their diffusion data (multiple scans). They saw correlations between a fiber-tract map they generated from the diffusion data (you plop down seed points in the cortex and other places and let fibers follow the main directions of diffusion) and their various activity correlation maps, and correlation was strongest for areas very close to each other on the brain and weak for longer fibers, especially inter-hemisphere fibers quite possibly because the tractography has a harder time getting those. The diffusion data also tended to show denser connections for stronger functional correlations, though as we see the instantaneous state can change the activity correlation quite a bit even though the white matter fiber tracts are not going to change that much on fast timescales. The fact that correlations are different in different activities illustrates that you dont need to have day to day changes entirely in the gross physical structure that shows up on scans of this type.
The actual layout of fibers at this coarse layer of detail is one thing of several that would contribute to activity correlations, including chemistry and actual engagement of said tract fibers for that particular activity and in that particular state, and all the fine molecular twiddling and potentiation at synapse scales.
Not only is data mining still going on by the group who published the paper, but Russ Poldrack (first author and subject of the study) is a very vocal proponent of open science: data associated with this publication have been made freely available for anyone else as well: openfmri.org
Also see this blogpost where he discusses creation of an open analysis platform (and the challenges in setting up analysis pipelines that are reproducible by others
Extreme Self-Tracking
Man has himself MRIed twice a week for a year and a half, plus tracking a lot about his life. The data mining is still going on, but at least it’s been shown that (probably) people’s connectomes change pretty rapidly.
I’m also posting this to the media thread because I’m not sure where it’s more likely to be seen.
Calling what was mapped here a ‘connectome’ is REALLY stretching it. When they make those graphs of parcels connected to each other, what they’re doing is just measuring the correlation between activity as revealed by an fMRI (which is itself removed from activity, measuring the short-term fluctuations in bloodflow as a result of energy requirements) in different parcels of brain and drawing a ‘connection’ when the coefficient is high enough. Correlation is not just connection.
I do note that there was diffusion tensor imaging (which shows you the average orientation of fibers in any given voxel [and showed an unusual crossing mixed fiber feature in a spot of his corpus callosum and will probably show similar oddities throughout the brain in any given human] ) and I will try to get at that information once I am past a paywall later on, but the repeated MRIs appear to be fMRIs.
I don’t think the paper is paywalled: link
MRIs: a lot of different scans including 15 T1 and T2 weighted structural scans; 19 diffusion-weighted scans. fMRI was mostly resting state (100), but also included various tasks such as n-back (15x), motion discrimination/stop signal (8x), object localiser (8x), verbal working memory localiser (5x), spatial WM (4x), breath holding (18x).
Oops! In whatever case I’m back from my situation-with-effectively-only-mobile-internet-access and have the paper now.
The repeated scans were indeed fMRIs measuring correlation of metabolic activity (a good proxy for activity) under various conditions. They made one diffusion tensor map from all their diffusion data (multiple scans). They saw correlations between a fiber-tract map they generated from the diffusion data (you plop down seed points in the cortex and other places and let fibers follow the main directions of diffusion) and their various activity correlation maps, and correlation was strongest for areas very close to each other on the brain and weak for longer fibers, especially inter-hemisphere fibers quite possibly because the tractography has a harder time getting those. The diffusion data also tended to show denser connections for stronger functional correlations, though as we see the instantaneous state can change the activity correlation quite a bit even though the white matter fiber tracts are not going to change that much on fast timescales. The fact that correlations are different in different activities illustrates that you dont need to have day to day changes entirely in the gross physical structure that shows up on scans of this type.
The actual layout of fibers at this coarse layer of detail is one thing of several that would contribute to activity correlations, including chemistry and actual engagement of said tract fibers for that particular activity and in that particular state, and all the fine molecular twiddling and potentiation at synapse scales.
Not only is data mining still going on by the group who published the paper, but Russ Poldrack (first author and subject of the study) is a very vocal proponent of open science: data associated with this publication have been made freely available for anyone else as well: openfmri.org
Also see this blogpost where he discusses creation of an open analysis platform (and the challenges in setting up analysis pipelines that are reproducible by others