Connectome

Overview of Connectome

  • Connectome Mapper 3 pipelines use a combination of tools from
    well-known software packages, including FSL, FreeSurfer, ANTs,
    MRtrix3, Dipy and AFNI, empowered by the Nipype dataflow library.
  • Connectome is a mind-bending adventure story that presents a daring scientific and technological vision for understanding what makes us who we are, both as individuals and as a species.
  • Connectome Mapper 3 implements full anatomical, diffusion and
    resting-state MRI processing pipelines, from raw Diffusion / T1 /
    T2 / BOLD data to multi-resolution connection matrices.
  • A connectome is a network map of effective synaptic connections and neural projections that comprise a nervous system and that shape its global communicative functions.
  • Connectomes are dynamic in that they are modified by what Seung (2012) refers to as the four Rs: ‘reweighting, reconnection, rewiring, and regeneration’ (p.
  • Connectome tells the incredible story of how Seung and a dedicated group of researchers are mapping these connections, neuron by neuron, synapse by synapse.
  • connectome was carried out by aggregating neighboring voxels according to a Voronoi diagram based on Euclidean distance between neighboring voxels (fig.
  • Connectome is an R toolkit to explore cell-cell connectivity patterns based on ligand and receptor data in heterogeneous single-cell datasets.
  • Connectome is a page turner—a book that should be read by anyone who lays claim to be thinking about the nature of life.”
  • A connectome is also a snapshot in time, which could reflect that particular worm’s past experience or specific traits.
  • Mine

    “This is complicated stuff, and it is a testament to Dr.Seung’s remarkable clarity of exposition that the reader is swept along with his enthusiasm, as he moves from the basics of neuroscience out to the farthest regions of the hypothetical, sketching out a spectacularly illustrated giant map of the universe of man.”—Abigail Zuger, M.D., New York TimesEvery person is unique, but science has struggled to pinpoint where, precisely, that uniqueness resides.Our genome may determine our eye color and even aspects of our character.But our friendships, failures, and passions also shape who we are.The question is: how?Sebastian Seung is at the forefront of a revolution in neuroscience.He believes that our identity lies not in our genes, but in the connections between our brain cells—our particular wiring.Seung and a dedicated group of researchers are leading the effort to map these connections, neuron by neuron, synapse by synapse.It’s a monumental effort, but if they succeed, they will uncover the basis of personality, identity, intelligence, memory, and perhaps disorders such as autism and schizophrenia.“Accessible, witty, imminently logical and at times poetic, Connectome establishes Seung as an important new researcher, philosopher and popularizer of brain science.It puts him on par with cosmology’s Brian Greene and the late Carl Sagan.”—Cleveland Plain Dealer

    Network

    The term “connectome” is commonly taken to describe a complete map of neural connections in a nervous system of a given species.This chapter provides a critical perspective on the role of connectomes in neuroscientific practice and asks how the connectomic approach fits into a larger context in which network thinking permeates technology, infrastructure, social life, and the economy.In the first part of this chapter, we argue that, seen from the perspective of ongoing research, the notion of connectomes as “complete descriptions” is misguided.Our argument combines Rachel Ankeny's analysis of neuroanatomical wiring diagrams as “descriptive models” with Hans-Jörg Rheinberger's notion of “epistemic objects,” i.e., targets of research that are still partially unknown.Combining these aspects we conclude that connectomes are constitutively epistemic objects: there just is no way to turn them into permanent and complete technical standards because the possibilities to map connection properties under different modeling assumptions are potentially inexhaustible.In the second part of the chapter, we use this understanding of connectomes as constitutively epistemic objects in order to critically assess the historical and political dimensions of current neuroscientific research.We argue that connectomics shows how the notion of the “brain as a network” has become the dominant metaphor of contemporary brain research.We further point out that this metaphor shares (potentially problematic) affinities to the form of contemporary “network societies.” We close by pointing out how the relation between connectomes and networks in society could be used in a more fruitful manner.

    Are there any recent connectome studies I could read?

    Sure thing.This study, published in September and based on data from the HCP, revealed a strong relationship between positive behavior traits and the wiring of your brain.While some brains appear to be wired for a lifestyle that includes education and high levels of satisfaction, other brains appear to be wired for anger, rule-breaking, and substance use, according to the Oxford researchers.

    How do we Map the Human Connectome?

    The Allen Institute for Brain Science in Seattle mapped the neural connections of a mouse brain.Five electron microscopes continuously ran for five months and collected over 100 million images of 25,000 slices of the mouse’s visual cortex.Each slice was 40 nanometers thick.By the end of their research, they collected 1.8 petabytes of data (that’s equivalent to 24 years of HD footage).Then, the institute developed a software program to assemble the images into a 3D volume and created a connectome.

    In the end, what will be gained from the project?

    In a previous email exchange about connectomics, Dr.

    Part 3: Brain Connectomics?

    00:00:07.17 Hi, I’m Jeff Lichtman,
    00:00:09.18 and I’d like to talk about
    00:00:12.17 tracing out wiring diagrams in the brain.00:00:14.29 In previous discussions,
    00:00:16.26 I’ve talked about the wiring diagram in muscle,
    00:00:21.04 where things are pretty straightforward.00:00:23.17 Still it’s not trivial to do these wiring diagrams,
    00:00:27.07 but except for me and a small number of people,
    00:00:31.13 most people are interested in the brain
    00:00:34.04 more than they are in muscle.00:00:36.00 I am one of those people who would
    00:00:38.06 prefer to study muscle to the brain,
    00:00:41.06 but clearly we would like eventually
    00:00:45.26 to be able to do wiring diagrams in the brain
    00:00:48.02 and do connectomics in the brain.00:00:50.20 It’s a challenge because the tools that we have previously developed,
    00:00:54.25 these tools that are called Brainbow
    00:00:56.24 where each nerve cell is labeled a different color,
    00:00:59.13 although they provide very nice-looking pictures of the brain
    00:01:03.09 — this is cerebral cortex, for example —
    00:01:07.14 it is hard to deal with the tracing of wires.00:01:10.24 If we zoom up on a small part of this piece of brain,
    00:01:14.03 this is sort of a Technicolor Golgi,
    00:01:17.02 that is the cells are many different colors,
    00:01:20.16 but what really matters is not the color of the cells,
    00:01:22.24 it’s the felt work,
    00:01:24.03 this fine stuff in between the cells,
    00:01:26.23 whether we can trace out those wires
    00:01:28.25 to see how one cell is connected to another.00:01:31.23 And with at least the original generation Brainbows,
    00:01:34.11 when we tried to use the best optical imaging techniques we could
    00:01:39.05 at the highest resolution possible
    00:01:41.17 — that’s Nyquist oversampling —
    00:01:43.22 the data
    00:01:45.00 — and using the highest numerical aperture lenses,
    00:01:47.11 for the aficionados —
    00:01:49.16 if you view an area like that
    00:01:51.20 at high resolution it just,
    00:01:54.03 not surprisingly perhaps,
    00:01:55.28 it’s just way too many wires there.00:01:58.05 It’s very hard to sort them out
    00:02:00.03 and one of the reasons is that within a single optical section,
    00:02:03.21 the thinnest you can focus,
    00:02:05.19 there are often multiple wires superimposed on top of each other,
    00:02:09.04 and you have a hard time,
    00:02:11.04 especially if the colors are hard to discriminate,
    00:02:13.24 and there’s only so many colors you get out of Brainbow,
    00:02:17.09 to know whether axons are crossing
    00:02:18.20 or just getting near each other and veering off.00:02:21.12 So the cell bodies are easy to see
    00:02:23.02 but the wires are a little problematic.00:02:25.29 If you want to look at a subset of cells
    00:02:28.06 in the central nervous system,
    00:02:30.10 Brainbow looks like a good technique,
    00:02:32.17 and one of the things Dawen Cai has done
    00:02:35.23 is develop a kind of virus that allows us
    00:02:40.17 to infect brains with a Brainbow set of viruses actually,
    00:02:46.22 that allow one to see very clearly subsets of cells.00:02:50.20 These are AAV viruses
    00:02:52.20 and if one has CRE only in a subset of cells in the brain,
    00:02:56.19 for recombination only to occur in that subset of cells,
    00:02:59.24 those cells will undergo color transformations and recombination;
    00:03:03.29 the other cells will remain dark.00:03:06.04 I’ll just show you one example of this,
    00:03:08.11 which is a piece of cortex
    00:03:12.03 in which the virus was used to infect cells
    00:03:17.25 that also have parvalbumin-driven CRE,
    00:03:21.24 and PV – parvalbumin is found in inhibitory neurons.00:03:25.02 What you’re seeing in this nest of axons
    00:03:28.00 are axons that surround the cell bodies,
    00:03:30.29 these dark objects like that,
    00:03:34.26 that are pyramidal neurons,
    00:03:37.05 and those pyramidal neurons are encrusted with inhibitory axons.00:03:40.23 And we think techniques like this
    00:03:43.06 will allow us to trace out, for example, inhibitory networks.00:03:46.02 You can see how much less dense this image is
    00:03:48.17 than the previous one.00:03:50.12 But what if we really want to see all the connections,
    00:03:53.16 and for many of the questions I am interested in personally,
    00:03:56.26 I really need to see all the connections.00:03:59.09 And that requires going to techniques
    00:04:01.26 that allow us to cut brains thin enough
    00:04:04.14 that every single section is disambiguous
    00:04:08.08 from the section in front and the section in back.00:04:11.27 When you use a confocal microscope,
    00:04:14.02 there’s not much you can be below about 3/4 of a micron.00:04:17.27 Things that are within 750 nm of each other
    00:04:20.25 tend to be in the same focal plane
    00:04:22.24 and it becomes very hard to disambiguate
    00:04:25.06 one thing from another.00:04:26.27 But if you could cut brains thinner than that,
    00:04:28.17 and I’m thinking about cutting brains
    00:04:30.18 not at 750 nm but 30 nm,
    00:04:33.21 then every wire is in its own section, basically,
    00:04:36.04 and that makes the reconstruction much better.00:04:41.03 So this is a tool that was built by Ken Hayworth
    00:04:46.09 with the help of Richard Schalek
    00:04:48.29 that replaces the human EM technician,
    00:04:54.22 the expert who knows how to take sections off with a microtome,
    00:04:58.26 this is the microtome here.00:05:01.06 This tool is an automatic collector of sections,
    00:05:05.26 and unlike a human being who uses an eyelash
    00:05:08.14 to push a section onto a thin film,
    00:05:12.27 this is a tape collecting device
    00:05:15.01 where the sections are collected on an actually quite sturdy,
    00:05:17.26 thick piece of plastic film.00:05:20.04 It’s an automated collection approach
    00:05:22.23 so we can collect up to 10,000 or so sections a day now,
    00:05:28.02 and because it’s done automatically
    00:05:30.07 and the machine never stops,
    00:05:32.09 we lose very few sections, we don’t destroy sections, they’re sturdy,
    00:05:36.13 and once it’s made these sections last a long time.00:05:40.18 A diagram of this device is shown here:
    00:05:42.16 it kind of looks like a movie projector,
    00:05:45.05 but it’s really a movie projector in reverse,
    00:05:47.10 in the sense we’re generating a film,
    00:05:49.17 these are the individual sections
    00:05:51.12 that are coming off of a piece of brain.00:05:53.18 And let me just zoom up on the guts of this, the important part,
    00:05:57.05 which is the collection part right up there.00:05:59.28 So here is a block of brain
    00:06:01.28 where the brain has been embedded in a very hard resin,
    00:06:06.23 and then a microtome is moving that piece of plasticized brain
    00:06:12.06 up and down against a diamond knife,
    00:06:15.05 and the diamond knife peels off a section
    00:06:17.23 that’s about 30 nm thick,
    00:06:19.23 or thinner than 30 nm often,
    00:06:22.20 which floats on water
    00:06:24.22 and then is picked up by this conveyer belt,
    00:06:26.24 and one section after another
    00:06:28.11 move back up on this conveyer belt.00:06:31.18 And each section is 30 nm in front of the section
    00:06:36.19 that came after it,
    00:06:38.07 and 30 nm on the other side of the next section.00:06:42.11 Once you have all the sections,
    00:06:43.23 you basically have the brain on a tape,
    00:06:46.21 and now this is a linearized version of the brain,
    00:06:50.28 and then we cut the tape up into strips
    00:06:54.21 and put the strips on a flat silicon wafer
    00:06:57.04 — you can see the sections on the wafer —
    00:06:59.27 and then image it in the electron microscope.00:07:04.09 Now, to generate a whole brain,
    00:07:06.20 or a large piece of brain,
    00:07:08.13 one has to make a whole library of sections like this
    00:07:11.20 and this is the idea here.00:07:13.23 One makes a library by taking the strips,
    00:07:17.04 putting them on a silicon wafer
    00:07:18.27 — this idea of a library was one of Ken Hayworth’s ideas
    00:07:21.24 when he built this in the first place —
    00:07:25.10 and that one wafer after another,
    00:07:28.16 you just keep going until you have the entire brain on wafers.00:07:31.17 And now we have a dataset of 55 such wafers, they’re about the size of CDs,
    00:07:38.09 which is a substantial chunk in the thalamus, for example.00:07:43.11 If we look at one of these wafers
    00:07:46.11 in the electron microscope
    00:07:48.20 that’s what it looks like:
    00:07:50.10 each of the sections, you can see one after another.00:07:54.24 We use little grids, the old-fashioned grids
    00:07:57.28 that were used to pick up sections before,
    00:08:00.19 now as fiduciary marks
    00:08:02.18 to allow us to align the wafer
    00:08:04.23 in the electron microscope.00:08:07.28 And if we zoom in one of these sections
    00:08:10.02 and look at that in the electron microscope
    00:08:12.15 that’s what a section might look like.00:08:14.26 So what you see is a bunch of blood vessels,
    00:08:17.10 these big light objects are blood vessels,
    00:08:20.18 there’s one up there
    00:08:22.11 and there are some smaller ones here,
    00:08:25.20 and then in between these gray circles,
    00:08:27.14 those are neurons,
    00:08:29.05 and these dark streaks
    00:08:31.14 are clusters of myelinated axons
    00:08:34.15 that are ensheathed in a lot of myelin.00:08:37.20 This is stained with a heavy metal of osmium
    00:08:40.04 that coats all the membranes dark,
    00:08:42.27 and myelin has a lot of membrane in it,
    00:08:44.18 so you see a lot of darkness.00:08:46.15 And then, again, between the cells
    00:08:48.04 all you have is felt work,
    00:08:50.08 and I’ll show you what’s in there in a moment.00:08:54.14 To see that felt work, we image this
    00:08:57.19 — and this is about 1 mm x 1mm image —
    00:09:01.23 at high resolution.00:09:03.03 So a light microscope, you have about the best resolution,
    00:09:06.29 if you don’t use super-resolution techniques,
    00:09:09.14 is about 250 or so nm, a quarter of a micron.00:09:14.00 We’re imaging this yellow box at 4 nm lateral resolution,
    00:09:21.00 and this section is 30 nm thick.00:09:25.01 So what does that section look like at 4 nm resolution?
    00:09:29.02 So here is that section at 4 nm resolution:
    00:09:31.26 it doesn’t look very impressive, it looks sort of like what I showed you before,
    00:09:34.27 but this is clearly not the full size image.00:09:38.03 This is the whole dataset,
    00:09:40.14 but in fact at 4 nm resolution,
    00:09:43.17 it’s 100,000 pixels from left to right
    00:09:46.00 and 100,000 pixels from top to bottom.00:09:48.04 This screen is about 1,000 pixels wide, roughly…00:09:52.03 roughly in that range.00:09:54.16 So this image is actually 100 times larger
    00:09:57.03 than the picture you’re looking at,
    00:09:59.02 and 100 times taller and wider.00:10:01.20 In fact, to see the whole picture,
    00:10:03.14 you’d have to back up so far it would look like this again.00:10:06.13 So I can’t show you the entire dataset at once.00:10:10.07 This image is a 10 billion pixel image,
    00:10:13.15 a 10 gigapixel image.00:10:15.11 Some of you have cameras that are 10 megapixel,
    00:10:18.06 this is an image that’s 1,000 times bigger,
    00:10:20.07 so you have to take 1,000 pictures for each of these,
    00:10:23.10 and that is because it’s a montage of 16 images,
    00:10:27.15 that’s what these streaks that you see are.00:10:29.24 It’s a 4 x 4 array of 25,000 x 25,000 pixel images,
    00:10:35.07 all knitted together to make this 100,000 x 100,000 image.00:10:39.04 That’s one image,
    00:10:40.19 and then we do that 10,300 times.00:10:43.19 So, huge dataset,
    00:10:45.19 and it’s about 100 terabytes,
    00:10:48.04 and we do about a terabyte per day
    00:10:51.02 as we take these images with our present tools.00:10:54.27 This is a single image,
    00:10:56.06 which gives you an impression of how much data there is in one image.00:10:58.29 I’d like to show you all the images,
    00:11:00.24 all 10,300 of them at once,
    00:11:03.02 gotta make them very small to see them,
    00:11:04.27 but there’s a point I’d like to show by doing that.00:11:07.27 And this is the 10,300 sections,
    00:11:10.20 there’s more than 100 TB of data,
    00:11:12.16 and each of these little boxes is one 10 GB
    00:11:15.06 — gigapixel —
    00:11:16.21 image,
    00:11:18.15 and one of the things you can see is that the quality of the machines
    00:11:21.01 we’re using require that they work
    00:11:24.05 minute after minute, 24 hours a day, 7 days a week,
    00:11:27.11 and never fail
    00:11:29.10 and, in fact, when you’re taking that much data,
    00:11:31.12 occasionally they do.00:11:33.03 Notice for example, this image here,
    00:11:35.25 and that image there
    00:11:38.12 and maybe that image there,
    00:11:41.07 where there’s no image.00:11:42.26 And this is because occasionally the machine
    00:11:44.24 would shut down for inexplicable reasons.00:11:47.08 When we talked to the company
    00:11:48.27 that made this microscope,
    00:11:50.10 they said, “Well, what do you expect?
    00:11:51.29 This is a crazy amount of data!
    00:11:54.11 We’ve never tested a machine for such long periods.”
    00:11:57.01 Notice here also that the images
    00:11:59.07 got progressively darker
    00:12:01.00 and in another place down here they got progressively lighter over time.00:12:04.27 The machines are not yet designed/optimized
    00:12:07.18 for this kind of continuous imaging.00:12:09.18 Eventually these problems will be worked out.00:12:12.16 So what do you get when you take all the images
    00:12:14.12 and stack them up?
    00:12:15.27 You generate a 100 TB dataset,
    00:12:19.10 in this case of the thalamus.00:12:21.17 It’s a lot of data and in some ways very impressive.00:12:25.06 In another way, not so impressive,
    00:12:27.23 if you actually look at the size of that dataset,
    00:12:30.12 it’s about the size of a grain of salt.00:12:33.08 You could take this data with a grain of salt, if you will,
    00:12:36.12 but it took us a long time,
    00:12:38.01 Josh Morgan did all the work of taking this image data,
    00:12:41.12 and that image data
    00:12:43.08 is really a tremendous repository now,
    00:12:45.04 because at that resolution that we’ve taken,
    00:12:47.04 we can see every synaptic vesicle in every synapse
    00:12:50.07 in something that you might think is small,
    00:12:52.22 but for us that’s a pretty large amount of data.00:12:56.15 The next thing I’d like to show is,
    00:12:58.06 what does the data actually look like?
    00:13:00.24 I’m gonna show you not a piece of thalamus,
    00:13:02.23 but a piece of cortex,
    00:13:04.16 by zooming the way you might in Google Maps
    00:13:06.27 start from a long distance
    00:13:08.22 and zoom in to finally see the structural details.00:13:12.18 And this is Bobby Kasthuri,
    00:13:14.09 who is the person who optimized
    00:13:15.27 our use of scanning electron microscopes for this.00:13:18.11 I asked him to hold his hand really still
    00:13:20.06 as we zoomed in on one of these 30 nm sections
    00:13:23.04 of cerebral cortex.00:13:24.27 So this is the cortex up here
    00:13:26.29 and the hippocampus is down below,
    00:13:29.13 and at some point he’s gotta enter the electron microscope,
    00:13:32.25 he obviously doesn’t do that himself
    00:13:34.20 but he puts the section in there.00:13:36.26 These are blood vessels and these are the nerve cells
    00:13:39.03 and these white streaks that you’re seeing running diagonally,
    00:13:42.18 those are the apical dendrites.00:13:45.01 These black circled objects are the myelinated axons,
    00:13:49.00 and these dark objects inside cells are mitochondria,
    00:13:52.12 and here is a synapse on a dendritic spine:
    00:13:54.26 a vesicle-filled profile of an axon
    00:13:57.01 making a synapse on a spine, with a spine apparatus
    00:14:00.09 that’s attached to this dendrite over here.00:14:03.24 That’s one section,
    00:14:05.07 but of course the idea here is to get a wiring diagram,
    00:14:07.28 we’re gonna have to do that section after section.00:14:11.01 So here is at low resolution for us, 30 nm per pixel,
    00:14:15.05 a series of several thousand sections
    00:14:17.09 that we’re going through.00:14:18.27 And you see cell bodies being cut,
    00:14:21.02 like this big object here,
    00:14:22.25 and that’s the nucleus of the cell body,
    00:14:24.22 and all the stuff moving around,
    00:14:26.29 and obviously this is from plastic, nothing is actually moving,
    00:14:30.28 this is not a time lapse image,
    00:14:32.26 it’s a space lapse image.00:14:34.26 We’re going from one section to the next,
    00:14:36.19 and as the wires pass through the sections,
    00:14:39.14 they appear to be moving from one place to another.00:14:42.11 And you can see at this low resolution lots of big objects,
    00:14:45.06 like these white objects that are moving around
    00:14:47.12 that are the big dendrites and these myelinated axons,
    00:14:50.13 and in between there’s other stuff,
    00:14:52.01 you don’t see it very well.00:14:53.15 The other thing you might notice is this image
    00:14:54.21 looks like it’s coming from 1903,
    00:14:58.16 it’s all grainy, and that’s because the tape underneath
    00:15:01.17 is giving a little bit of noise.00:15:03.05 This by the way is a blood vessel down here,
    00:15:05.01 that’s the endothelial cell and that’s the entrance
    00:15:07.25 where the blood cells are.00:15:09.21 If we zoom up at higher resolution
    00:15:11.22 and look at the same data,
    00:15:13.29 you see that between these large objects moving around
    00:15:16.29 are lots of little wires also moving around.00:15:19.23 The brain is just filled with wires,
    00:15:21.19 this is a big cell body of a neuron up here
    00:15:23.28 — pyramidal cell —
    00:15:25.18 and all these little wires are moving from one place to another,
    00:15:28.09 filled with vesicles,
    00:15:29.29 those vesicles are where the synapses are.00:15:32.02 If you look carefully,
    00:15:33.08 you’ll see that there are synaptic vesicle-filled profiles
    00:15:35.16 almost everywhere in the gray matter,
    00:15:38.00 and if you look carefully also you feel like
    00:15:40.15 you can follow many of these objects
    00:15:42.13 from one section to the next.00:15:44.25 In fact, if you gave a five-year old a coloring book
    00:15:48.28 with 1,000 or 10,000 pages of black and white like this,
    00:15:52.13 gave them a crayon
    00:15:54.01 and asked them to color the same object in
    00:15:56.03 in the same color from section after section,
    00:15:58.13 you could get the wiring diagram of data like this.00:16:01.29 Daniel Berger built such a coloring book for digital use,
    00:16:07.14 and I’ll show you that in the following slide.00:16:12.17 So this is a digital coloring book,
    00:16:15.13 where anyone who wants can go in and color in an object,
    00:16:19.10 and color it in the same color
    00:16:20.24 section after section after section.00:16:23.24 And you can see that it’s easy to follow these objects
    00:16:27.26 from one section to another,
    00:16:29.17 especially when you color them all in.00:16:31.06 The gray objects here were also colored in,
    00:16:33.05 but we kept them gray because they’re the glial cells,
    00:16:35.01 they’re not neurons.00:16:36.21 And from this data, you can then take,
    00:16:39.20 just generate the 3-dimensional version of it.00:16:41.24 This is the same data exactly,
    00:16:43.14 these are all those dendrites with their spines,
    00:16:45.22 and then packed in between them
    00:16:47.13 are lots of little axons that are making connections with them.00:16:51.00 And what’s let out are the glial cells,
    00:16:52.27 those are the little cavities you see,
    00:16:54.24 and there’s not that much glial stuff in there
    00:16:56.15 relative to neurophil, as it’s called.00:16:59.10 Now that looks impressive, but it’s really useless.00:17:02.15 It’s useless in large part
    00:17:03.29 because every side of this box is orphaned;
    00:17:07.17 we don’t know where it came from.00:17:09.05 We’re looking at a small region of brain,
    00:17:10.27 but how small is really what’s painful to look at here.00:17:14.13 So there is the region that was reconstructed
    00:17:16.28 in this section,
    00:17:18.19 and that section sits here,
    00:17:22.12 in this piece of a section,
    00:17:25.24 and this piece of a section sits there,
    00:17:27.26 in this whole section of cortex.00:17:31.15 So that little green blinking dot, if you can see it,
    00:17:35.17 which is the smallest dot I can make on this screen,
    00:17:38.09 is bigger than what we actually ended up in doing.00:17:42.29 And to make matters even worse,
    00:17:45.02 let me give you a sense of what a single voxel
    00:17:47.20 of an fMRI image is, that’s a cubic millimeter.00:17:51.01 By this standard,
    00:17:53.23 a cubic millimeter is that big.00:17:56.12 So what we’ve actually ended up doing
    00:17:58.15 is a very small amount of brain,
    00:18:01.13 and clearly not sufficient.00:18:03.17 Even taking the images,
    00:18:05.26 much less coloring them in,
    00:18:07.11 just taking the images takes time.00:18:09.04 If we wanted, let’s say, to do a cubic millimeter,
    00:18:12.04 at the rate we were going
    00:18:13.04 when we began doing this work
    00:18:16.23 over three years ago,
    00:18:18.24 we were doing around 0.5M pixels/second in our imaging,
    00:18:23.18 and that would require us to go about 2.24 centuries,
    00:18:27.24 just to generate the data for a cubic millimeter.00:18:31.01 And that would be about 2,000 TB,
    00:18:33.23 and no graduate student I met was interested in this project
    00:18:38.19 for obvious reasons,
    00:18:42.02 ’cause they knew I would be dead by the time they finished,
    00:18:44.27 as well as them,
    00:18:46.18 so this was not possible
    00:18:49.13 to work at those speeds.00:18:51.28 So since then we’ve been optimizing the imaging techniques
    00:18:55.07 to go faster and faster.00:18:56.28 At the moment we go about 20M pixels/second,
    00:19:01.00 so 40 times faster,
    00:19:03.07 and we can do about 1 TB a day,
    00:19:06.01 and we can do a cubic millimeter of data in 5.6 years.00:19:09.25 We just have a new machine delivered
    00:19:13.07 that goes about 40 M pixels/second,
    00:19:16.23 and so eventually we hope to be doing
    00:19:19.15 a cubic millimeter in 2.8 years.00:19:21.26 That’s still too long to do I think useful,
    00:19:24.27 comparative biology of different brains
    00:19:29.12 that maybe have different experiences
    00:19:30.26 or different diseases,
    00:19:32.27 but soon we hope to be going much faster,
    00:19:35.23 another 60-fold faster or so,
    00:19:38.14 going over a billion pixels/second,
    00:19:40.28 and thinking about doing cubic millimeters
    00:19:43.20 in less than a month, maybe even a few weeks.00:19:46.08 And the way we’re gonna go that fast
    00:19:47.28 is by instead of using scanning electron microscopes
    00:19:50.01 that have a single beam,
    00:19:51.17 using scanning electron microscopes
    00:19:53.06 that each have many beams in them.00:19:55.17 In collaboration with Zeiss,
    00:20:01.06 a company that is building this in Germany,
    00:20:03.22 they have been building a 61 beam scanning electron microscope,
    00:20:08.02 which gives us 61-fold more beams
    00:20:11.09 scanning at any one time.00:20:12.24 It’s like having 61 microscopes all in one.00:20:15.14 The original prototype is a hefty looking device,
    00:20:18.09 especially when you compare it to the size of a normal-sized human being.00:20:22.11 It’s a massive and impressive piece of technology,
    00:20:26.20 and German engineering really comes to the fore in devices like this.00:20:31.08 So we hope eventually to be going fast enough
    00:20:34.07 to do serious biology with connectomes.00:20:37.13 Of course, that just gives you the images,
    00:20:39.00 how about tracing it?
    00:20:40.15 You know, the tracing is a slow process.00:20:43.00 The area we traced before
    00:20:44.10 is just this miniscule part of this image.00:20:46.13 What you’d like is to have everything colored in,
    00:20:48.19 and here is everything colored in,
    00:20:50.15 but not colored in by a human being,
    00:20:52.08 but colored in by a machine, by computers.00:20:54.14 This is work
    00:20:55.22 that we’re doing in collaboration with an engineering lab at Harvard,
    00:20:59.21 Hanspeter Pfister’s lab.00:21:01.13 And that kind of data
    00:21:03.17 generates wiring diagrams
    00:21:06.22 much, much faster
    00:21:08.19 than humans can generate them.00:21:10.14 So this is entirely automated,
    00:21:12.14 and the algorithm is getting better and better and better.00:21:16.01 If you look carefully at this movie,
    00:21:17.13 you might find an object that changes color.00:21:19.21 If it does, the algorithm failed us;
    00:21:21.20 it means it got confused.00:21:23.26 Most recently,
    00:21:26.25 there are parts of the brain where you really have to look hard
    00:21:28.12 to find any errors at all,
    00:21:30.14 but we think this is the future
    00:21:32.08 for this kind of analysis,
    00:21:34.28 to do this without humans doing a lot of tracing.00:21:38.02 So what do we do with data like this?
    00:21:39.22 I’m just gonna end with a couple of movies
    00:21:41.16 to give you a sense of what we’re doing.00:21:43.12 So we sparsely reconstruct a bunch of neurons
    00:21:46.03 in a cortical column in the cerebral cortex
    00:21:48.18 and then we decided, for one of those neurons,
    00:21:51.00 we would learn everything we possibly could
    00:21:54.04 about 1,000 cubic microns
    00:21:57.09 around one apical dendrite
    00:21:59.18 of one pyramidal cell.00:22:01.18 So this little cylinder is everything inside
    00:22:04.08 one little region of the dendrite of one cell,
    00:22:07.11 just around it,
    00:22:09.13 and in the sense, because it’s all surrounding that dendrite,
    00:22:12.05 every axon in there has a potential
    00:22:14.12 to innervate that particular dendrite.00:22:17.06 But what else is in there?
    00:22:18.25 It turns out there’s a lot of stuff in that little region.00:22:22.10 There’s about 675 synapses,
    00:22:25.28 530 different axons,
    00:22:27.28 90 different dendrites,
    00:22:29.29 and when we look at the connectivity,
    00:22:32.02 it doesn’t look random at all,
    00:22:35.09 it looks, in an interesting way,
    00:22:37.13 that some axons prefer to make multiple synapses on some dendrites,
    00:22:42.03 and other axons prefer to make multiple synapses on other dendrites.00:22:46.09 You can see that by tracing out a single axon,
    00:22:49.14 but I want to give you a sense of what it means
    00:22:51.13 to have 675 synapses in one little area like this,
    00:22:55.22 by showing you all the synaptic vesicles
    00:22:58.09 of all the synapses in that area.00:23:00.21 You just have this cloud of synapses around this dendrite,
    00:23:03.24 and many of those synapses
    00:23:05.13 are not on this dendrite.00:23:06.24 This dendrite only has 80 excitatory synapses on it,
    00:23:09.21 but there are about 600 synapses in that region.00:23:13.11 If we look at a single axon at a time,
    00:23:15.28 you see that many of these axons
    00:23:17.19 seem to have some special affinity for this dendrite,
    00:23:21.01 and other axons have a special affinity
    00:23:23.00 for other dendrites,
    00:23:24.17 even though they’re all excitatory axons.00:23:28.01 And then because, as I said,
    00:23:29.17 we can see individual synaptic vesicles,
    00:23:32.07 we can go in and look at the synaptic vesicles of every synapse
    00:23:36.03 of every axon innervating this cell,
    00:23:38.08 and I’ll just end by showing you some pictures like that,
    00:23:41.14 not so much to make a big point of
    00:23:45.26 what we have learned,
    00:23:47.24 but how much we don’t understand.00:23:50.21 The brain is vastly complicated.00:23:52.22 This is not an artist’s rendering of the brain.00:23:54.21 These yellow dots,
    00:23:56.16 each of them is a synaptic vesicle inside a synapse.00:24:00.06 When you look at the brain at this high level of resolution,
    00:24:03.23 and here’s one last view of this,
    00:24:06.26 I think you get the impression that the brain
    00:24:09.07 is extraordinarily complicated,
    00:24:12.14 but maybe at this level
    00:24:14.04 one will learn sort of the way information
    00:24:16.27 is instantiated in brain structure.00:24:19.20 One may learn what diseases like schizophrenia
    00:24:22.11 actually look like
    00:24:24.01 as a physical disease of the brain.00:24:26.11 And ultimately, I think
    00:24:27.28 you get a really good sense of humility,
    00:24:30.15 because at least my brain,
    00:24:32.01 what I do with my brain, the thoughts that come out of my brain,
    00:24:34.23 are far less impressive, if you will,
    00:24:38.23 than the complexity of the machine that allows me to do that thinking.00:24:42.18 I think I’ll end with that thought
    00:24:45.25 and just recognize the many, many people
    00:24:49.12 who did this work.00:24:50.28 I’ve been a cheerleader for a lot of this work,
    00:24:53.19 but I want to point out that this kind of science cannot, I don’t think,
    00:24:59.11 be done in a single laboratory.00:25:01.11 One needs computer scientists,
    00:25:03.10 one needs people who are very good with molecular biology,
    00:25:08.12 one needs engineers,
    00:25:10.02 and I’m grateful to a wide range of colleagues
    00:25:14.10 for much of the work I’ve talked about.00:25:16.13 Thank you.

    So this is it, or is there more?

    Not by a long shot.Commentary on the project suggests both the technology and methodology, even if it is the most advanced in existence, had limitations that may have biased interpretation of the brain’s architecture.That said, the connectome is a moving target — the more we learn, the more there is to learn.

    The Connectome Debate: Is Mapping the Mind of a Worm Worth It?

    Scientists have mapped a tiny roundworm's entire nervous system.

    The Connectome Debate: Is Mapping the Mind of a Worm Worth It?

    Scientists have mapped a tiny roundworm’s entire nervous system.

    What are the data types?

    Grayscale data correspond to traditional electron microscope images.This is written only once, after alignment, but often read, because it is required for segmentation, synapse finding, and proofreading.We store the grayscale data, eight bits per voxel, in Google buckets, which facilitates access from geographically distributed sites.

    What is a connectome?

    The suffix “-ome” refers to a totality, as The Scientist explains.So, just as a genome would refer to all of the DNA within a single cell, a connectome refers to all the cellular connections within the brain or, more simply, a connectome is a wiring diagram of the brain.In fact, each map reveals the architecture of our brain’s white matter.More specifically, each diagram shows how bundles of fiber course through the white matter of the brain to form functional connections between gray matter regions.The fiber bundles look like “colored spaghetti,” Dr.Arthur Toga, a neuroscientist at USC’s Laboratory of Neuroimaging, Keck School of Medicine told Medical Daily.

    What is Connectomics?

    Connectomics is the study of a connectome.A connectome is a complete map of your neural connections.Connectomes can also be maps of specific brain subsystems, such as the neural connections in your hippocampus.The goal of these wiring diagrams is to understand how the structure of the brain is connected to its behaviour.

    What kind of technology was used?

    Only the coolest.To conduct the work, scientists used noninvasive neuroimaging equipment, including customized head coils and innovative MRI hardware.The high-resolution tools and technology are not only more sensitive but they also require less time to produce an image, compared to the past.Along with advanced computer science, the technologies reveal the brain as a whole, and at a level of detail not previously imagined possible in a living person.

    What’s been discovered so far?

    Many surprises.Scientists have been amazed to see that, instead of chaos, the connecting fibers are organized into an orderly 3D grid, where axons run up and down and left and right, minus any diagonals or tangles.Science magazine compares the brain’s 3D layout to New York City, with its streets running in two directions and buildings’ elevators running up and down.Strangely, in flat areas of the grid, the fibers overlap at precise 90 degree angles and weave together much like a fabric, the scientists say.

    What’s Connectome?

    Connectome is a technology platform to realize human-like AI assistant, “Virtual Human Agent” (VHA).By harnessing the Artificial Intelligence (AI), Game AI (human-like intelligent NPC technology), blockchain and human sciences, VHAs can, amongst other things, be personal assistants, the cornerstone of productive organizations and companies, assist in healthcare, and be the future of human-technology interaction.The goal is to create a future where humans can trust in and live alongside AI technologies that will increase the quality of communication between humans, as well as between humans and technology.

    What’s the project’s timeline and did I miss it?

    The project was carried out in two phases.During Phase I, which spanned the years 2010 through 2012, research teams designed the project’s 16 major components.During Phase II, which ranged from 2012 through this past summer, the various scientists performed the actual work of gathering data.More importantly, however, during the most recent phase investigators made their datasets publicly available at regular intervals so that scientists around the world could begin to use them in their own projects.

    Who helped the scientists? Who were the participants?

    A total of 1,200 healthy adults, including a high proportion of twins and their non-twin siblings.This unusual emphasis on twins and their families was intended to help the researchers understand whether brain circuits are inherited.As part of their contribution, each participant had their genome mapped, allowing scientists to evaluate how much genes influence — or don’t influence — individual brain wiring.Each volunteer also completed a series of questionnaires and tests to measure behavior and demographic traits.

    Who is behind the Human Connectome Project?

    The project comprises 36 investigators, including biologists, physicians, physicists, and computer scientists, at 11 institutions across the nation.The primary centers of research are USC’s Laboratory of Neuroimaging, Massachusetts General Hospital’s Martinos Center, Washington University’s Van Essen Lab, and the University of Minnesota’s Center for Magnetic Resonance Research.

    Who’s Involved?

    “We have a chance to improve the lives of not just millions, but billions of people on this planet through the research that’s done in this BRAIN Initiative alone.But it’s going to require a serious effort, a sustained effort.

    Why EyeWire?

    Solving the mysteries of the brain requires something more powerful than a supercomputer – YOU.Together, we’re mapping neural circuits to decipher the mysteries of vision.

    History of Connectome

  • In the 1970s biologist Sydney Brenner and his colleagues began preserving tiny hermaphroditic roundworms known as Caenorhabditis elegans in agar and osmium fixative, slicing up their bodies like pepperoni and photographing their cells through a powerful electron microscope.
  • In 1986 the scientists published a near complete draft of the diagram.
  • In 2004, Harvard established the Center for Brain Science (CBS) on its Cambridge campus, an interdisciplinary Center aimed at understanding neural circuits—their structure, their function, and how they are changed during development, aging, and disease.
  • In 2004, researchers at the Max Planck Institute in Germany had demonstrated automated methods that could analyze images of neurons produced by electron microscopes—a process known as segmentation.
  • In 2005, Dr.
  • In 2005, Indiana University neuroscientist Olaf Sporns and his colleagues coined the term “connectome” to denote a comprehensive, structural blueprint of the human brain (1).
  • In 2012 and 2013, we broke down the top five most fascinating, transformative, implication-riddled neuroscience discoveries of the year.
  • In 2017, a first article under the lead of Andrea A Kühn made use of such normative connectomes by combining connectivity measures with electrophysiological recordings of the subcortex.
  • In the 1970s and 1980s, a team of researchers traced a map of its 7,000 interneural connections.
  • In the 1980s, as a postdoctoral student in Brenner's lab, Martin Chalfie—now at Columbia University—used the C.