[ExI] HM

[Anders Sandberg]

Eugene:
> I don't think a single scanner stage can handle more than a cube of cm^3 or
> so.
> Even that might be a tad on the optimistic side.

Yes, I think the KESM is the only system right now that can handle that. For Kenneth Hayworth's ATLUM I think the volume is many order of magnitude smaller. 

> > Todd and the KESM team have a 2 Tb datasets of a mouse brain, but that is
> of course just optical resolution. Enough to see the neurons, but not enough
> to get connectivity.
> 
> Right now we can handle about a PByte dataset comfortably. 10 PBytes less
> comfortably, and 100 PBytes with a lot of pain.

Even a terabyte dataset is a bit tricky to use, even if it fits nicely onto a single hard drive. For example, just doing a visualization or flythrough of that mouse brain seems to be a challenging software project. 

> for tracing/segmentation
> you have to work with a sliding slice (anywhere in a 1 um-100 
> um depth). The result would be a lot more compact, and also quite
> compressible (very little difference between adjacent slices).

Yup. What I worry about is how much sliding we can do between different lateral pieces of tissue. If we cannot use the same stage for an entire slice we will have to split it between stages. This has to be done *really* carefully so we can match the edges for connectivity. I don't know if there has been much work on how this can be done well - freeze cracking, maybe?


> The problem is that a dm^3 is 1000 cm^3, so these cm^3 scanner
> stages (at about ~nm resolution laterally) be better cheap. This 
> is something which will make even Paul Allen's people blanch.

The wonders of mass production. Right now microscope stages are sold one by one, and tend to be expensive (partially because they are generic for any preparation). 


Anders Sandberg,
Future of Humanity Institute
Philosophy Faculty of Oxford University