Plastination versus Cryonics

Break down survival as Drake equation, see how plastination differs from cryonics, try to calculate advantage
biology, transhumanism, survey
2011-07-242014-10-22 in progress certainty: likely importance: 9

“No man is an iland, in­tire of it selfe; every man is a peece of the Con­ti­nent, a part of the maine; if a clod bee washed away by the Sea, Eu­rope is the lesse, as well as if a Promon­to­rie were, as well as if a Man­nor of thy friends or of thine owne were; any mans death di­min­ishes me, be­cause I am in­volved in Mankinde; And there­fore never send to know for whom the bell tolls; It tolls for thee….”

, Med­i­ta­tion 17

The for is just a num­ber of se­quen­tial steps with in­de­pen­dent prob­a­bil­i­ties, all of which must suc­ceed but oth­er­wise none more im­por­tant than the oth­ers. (One soft­ware ver­sion is the Cry­on­ics Cal­cu­la­tor.) Web de­vel­op­ers call it a . Spe­cific equa­tions and val­ues have been pro­posed, usu­ally yield­ing prob­a­bil­ity of suc­cess 0<x < 10%. For ex­am­ple, Steven Har­ris in 1989 es­ti­mated 0.2-15%, R. Mike Perry in the same ar­ti­cle runs a differ­ent analy­sis to ar­rive at 13-77%, Ralph Merkle sug­gests >85% (con­di­tional on things like good preser­va­tion, no dystopia, and nan­otech); Robin Han­son cal­cu­lated in 2009 a ~6% chance, Roko gave 23%; Mike Dar­win in 2011 (per­sonal com­mu­ni­ca­tion) put the odds at <10%; an in­for­mal sur­vey of >6 peo­ple (LW dis­cus­sion) av­er­aged ~17% suc­cess rate; Jeff Kauf­man in 2011 pro­vides a cal­cu­la­tor with sug­gested val­ues yield­ing 0.2%; The 2012 Less­Wrong sur­vey yields a mean es­ti­mate of cry­on­ics work­ing of 18% (n = 1100) and among ‘vet­er­ans’ the es­ti­mate is a lower 12% (n = 59) - but in­ter­est­ing­ly, they seem to be more likely to be signed up for cry­on­ics.


One such Drake equa­tion might break out the steps as fol­lows:

  1. Like­li­hood of get­ting pre­served
  2. * preser­va­tion con­tains needed in­for­ma­tion
  3. * in­for­ma­tion’s sur­vival over the cen­turies un­til re­vival pos­si­ble
  4. * ex­is­tence of or­ga­ni­za­tions or en­ti­ties ar­rang­ing re­vival
  5. * the ac­tual re­vival

With those 5 val­ues, one mul­ti­plies to get the fi­nal prob­a­bil­ity of each step com­ing true and hence of a suc­cess­ful re­vival. Be­cause each step is mul­ti­plied to­gether with no weights, im­prove­ments are equal - an im­prove­ment in one fac­tor is as good as the same im­prove­ment in an­other fac­tor: a 10% im­prove­ment in or­ga­ni­za­tional con­ti­nu­ity is as good as a 10% im­prove­ment in the odds that the vit­ri­fi­ca­tion pre­serves nec­es­sary in­for­ma­tion, which is as good as a 10% im­prove­ment in odds that re­vival tech will be de­vel­oped. This also holds for bal­anc­ing profit and loss (it’s all the same). A tech­nol­ogy that in­creases the or­ga­ni­za­tional pa­ra­me­ter by 30% and de­creases the in­for­ma­tion preser­va­tion pa­ra­me­ter by 10% would be a net gain, be­cause the gain in one step out­weighs the loss in an­oth­er, re­gard­less of what con­crete val­ues one as­signs. (For ex­am­ple, if X was 50% and Y was 60% for a fi­nal chance of 30%, then you would be bet­ter off if you could do some­thing differ­ent where X was 80% and Y was 40% be­cause that yields a fi­nal chance of 32%. This would be eas­ier to see in differ­ent no­ta­tion like .)


Bi­o­log­i­cal sam­ples have been ac­ci­den­tally pre­served from the deep past through de­hy­dra­tion, freez­ing, anox­ia, and chem­i­cal preser­va­tion; has (pos­si­bly) been re­cov­ered from 250 mil­lion year old salt crys­tals, 23 mil­lion year old in­sects are clas­si­fi­able and pre­served in high fi­deli­ty, and ice sam­ples have pre­served 800,000 year old and 400,000 year old DNA. Some re­cov­ery has been ac­com­plished of 400,000 year old ho­minid DNA, 45,000 year old hu­man and 38,000 year old Ne­an­derthal DNA has been , as has 28,000 year old woolly mam­moth DNA, 80,000 year old ho­minid DNA, and DNA. 30,000 year old frozen plant tis­sue has been grown into healthy adult plants. One 4000 year old hu­man genome was se­quenced. A 2700 year old hu­man brain has been re­cov­ered from a wa­ter­logged Eng­lish pit, heav­ily dam­aged but vis­i­bly still a brain; it is one of hun­dreds of brains re­cov­ered from wa­tery en­vi­ron­ments. Many of the sam­ples chem­i­cally pre­served in turned out to be , but nev­er­the­less, the is very good and down to the cel­lu­lar level:

Ex­am­i­na­tion of the ul­tra­struc­ture of pre­served tis­sue in the ab­domen of a fos­sil fly (Myce­tophil­i­dae Diptera) en­tombed in Baltic am­ber re­vealed rec­og­niz­able cell or­ganelles. Struc­tures that cor­re­sponded to mus­cle fibers, nu­clei, ri­bo­somes, lipid droplets, en­do­plas­mic retic­u­lum, and mi­to­chon­dria were iden­ti­fied with the trans­mis­sion elec­tron mi­cro­scope. Preser­va­tion was at­trib­uted to in­ert de­hy­dra­tion as well as the pres­ence of com­pounds in the orig­i­nal sap which func­tioned as nat­ural fix­a­tives. This ev­i­dence of cell or­ganelles in fos­silized soft tis­sues rep­re­sent an ex­treme form of mum­mi­fi­ca­tion since Baltic am­ber is con­sid­ered to have formed about 40 mil­lion years ago.

(Even the color in di­nosaur feath­ers has been pre­served in am­ber.) Ben Best de­scribes am­ber’s preser­v­a­tive mech­a­nism in “An­cient DNA & Preser­va­tion in Am­ber”:

“Tree sap (resin) con­tains sug­ars as well as al­co­hols & alde­hy­des (in­clud­ing ter­pe­nes), which are de­hy­drat­ing & an­tibi­otic as well as pro­vid­ing an air-tight seal to pre­vent fur­ther en­try of oxy­gen. Myrrh is a mix­ture of resin, gum and es­sen­tial oils from the Com­miphora plant that was used by the an­cient Egyp­tians for em­balm­ing (by pour­ing it into the cra­nial, chest, ab­dom­i­nal and pelvic cav­i­ties) and mum­mi­fi­ca­tion (by soak­ing the wrap­ping ban­dages in it)….Am­ber, as a sticky pitch from cer­tain trees, can trap in­sects when fresh from a tree-wound. The sug­ars, al­co­hols & ter­pene-alde­hy­des diffuse into the in­sect to de­hy­drate & pre­serve. The am­ber sur­rounds the in­sect, pro­vid­ing an air-tight seal. Fur­ther ox­i­da­tion & poly­mer­iza­tion of the ter­penes pro­tect the in­sect from fur­ther dam­age. The con­tin­ued poly­mer­iza­tion of the am­ber ter­penes even­tu­ally re­sults in an in­sol­u­ble gem­stone-qual­ity glass that pre­serves the in­sect in a strong en­case­ment. Al­though such for­tu­itous com­bi­na­tion of chem­i­cal preser­va­tion and oxy­gen-tight en­case­ment should not be ex­pected for preser­va­tion of large spec­i­mens (like hu­mans or di­nosaurs), the use of some hard­ened plas­tic or resin en­case­ment could as­sist chem­i­cal and/or de­hy­dra­tion preser­va­tion.”

and chem­i­cal brain preser­va­tion have been se­ri­ously pro­posed1 as an al­ter­na­tive to cry­on­ics, ap­par­ently first by Charles Ol­son in “A Pos­si­ble Cure for Death”. R. Mike Perry dis­cusses it fa­vor­ably in “The Road Less Trav­eled: Al­ter­na­tives to Cry­on­ics”. Greg Jor­dan says no con­vinc­ing coun­ter-ar­gu­ments have been raised since Ol­son and strongly ap­proves of it in his post “Biosta­sis through chemo­p­reser­va­tion”.

Is plas­ti­na­tion a net gain?


Ad­van­tage for plas­ti­na­tion:

  1. Im­proves sur­vival pa­ra­me­ter #3: It is prob­a­ble that scan­ning tech­nol­ogy will out­strip up­load tech­nol­o­gy. In many fields, the abil­ity to gather data ex­ceeds the abil­ity to process or un­der­stand it. Hence, it is pos­si­ble and quite likely that dur­ing the long wait for re­vival, it will be­come pos­si­ble to scan a plas­ti­nated brain in suffi­cient res­o­lu­tion to even­tu­ally up­load it.

    Even if the scan were de­struc­tive, such a scan would make it pos­si­ble to dras­ti­cally in­crease sur­vival odds by copy­ing the dig­i­tal data to many archives and for­mats on­line and offline. No such op­tion is avail­able to a cry­on­ics brain un­less it aban­dons cry­on­ics en­tire­ly, in which case why did it take the risk of the cry­on­ics fail­ing & it warm­ing up rather than be plas­ti­nated from the be­gin­ning? It’s hard to imag­ine the ben­e­fits be­ing so equally bal­anced that the ac­tu­al­iza­tion of bet­ter scan­ning would is enough to change the plans - given how many pa­ra­me­ters there are, a ‘pure’ strat­egy of 100% cry­on­ics or 100% plas­ti­na­tion will win. (In­deed, one might won­der how one would know that a plas­ti­na­tion+s­can­ning pro­ce­dure was good enough for up­load­ing in the ab­sence of a suc­cess­ful hu­man up­load. Hu­man bi­ol­ogy often di­verges from even close an­i­mal mod­els, and should­n’t we ex­pect things like con­scious­ness to be even less re­li­ably mod­eled by those an­i­mal mod­els? The win­dow be­tween the first suc­cess­ful up­load and wide­spread up­load­ing will be short com­pared to the time be­tween now and then, even if you as­sume no Sin­gu­lar­ity of any kind, not even Robin Han­son’s Crack of a Fu­ture Dawn, and a slowed-down Moore’s law.)

  2. Im­proves or­ga­ni­za­tional pa­ra­me­ter #4: Plas­ti­na­tion may be such a tech­nol­o­gy. It does not re­quire or­ga­ni­za­tional con­ti­nu­ity; one rough year and your brain is a pile of rot­ting mag­gots with cry­on­ics. one rough year with plas­ti­na­tion, and your brain is a bit dusty2. A plas­ti­nated brain does­n’t even need an or­ga­ni­za­tion: it may be pre­served as a time cap­sule, a fam­ily heir­loom, a cu­rios­i­ty, or per­haps just buried some­where; but a cryo­geni­cally stored brain must have a so­phis­ti­cated sup­port sys­tem which will sup­ply it reg­u­larly with liq­uid ni­tro­gen, and that rules out pretty much every­one but a cry­on­ics or­ga­ni­za­tion. Mike Dar­win has been a real wake up call - the Out­side View says and are much more risky than usu­ally as­sumed3 - and in­deed, one cry­on­ics or­ga­ni­za­tion has al­ready failed with the loss of pa­tients. Past the cen­tury mark, a few per­cent­age points is the op­ti­mistic es­ti­mate! Cry­on­ics or­ga­ni­za­tions have done rea­son­ably well so far, but ALCOR con­sis­tently runs at a loss and if mem­ber­ship does not fol­low an ex­po­nen­tial growth (as it does not), then rel­a­tively soon the ra­tio of dead mem­bers to live mem­bers will start get­ting worse.

  3. Im­proves like­li­hood of preser­va­tion #1: Much cheaper than vit­ri­fi­ca­tion; while cryo­genic stor­age is very cheap in scale the cost is still non-triv­ial for the fore­see­able fu­ture.

  4. Im­proves re­vival pa­ra­me­ter #5:

    1. de­spite be­ing a rel­a­tively young field (al­beit re­spectable & well-fund­ed), plas­ti­na­tion & scan­ning has made tremen­dous progress and is slowly be­ing au­to­mated, with one hu­man brain sliced at 70 mi­cro thick­ness and pho­tographed4, or pro­duc­ing par­tial 5 of brains. One might char­ac­ter­ize the two fields as: con­nec­tome:u­pload­::re­vived-rab­bit-kid­ney:­func­tion­ing-brain, and pon­der the fol­low­ing pos­si­bil­i­ties:

      • (=) If one re­gards the ‘dis­tance’ be­tween the state of the art and the goal as equal, then plas­ti­na­tion’s faster progress is a win
      • (<) If one re­gards the dis­tance as smaller for plas­ti­na­tion than cry­on­ics, then plas­ti­na­tion wins both on faster progress and how much is left to do
      • (>) Only if one re­gards the kid­ney as be­ing much closer to a re­viv­ing a func­tion­ing brain can it be pos­si­ble for cry­on­ics re­vival to beat plas­ti­na­tion re­vival.

      Pon­der­ing the Roadmap and the Blue Brain pro­ject, I strongly doubt the kid­ney-brain is much closer to­gether than con­nec­tome-u­pload, and sus­pect that the lat­ter is clos­er.6

    2. If plas­ti­na­tion turns out to be the ‘right’ start­ing point for an up­load and cry­on­ics brains must be plas­ti­nated first, we might ex­pect the cry­on­ic­s->­plas­ti­na­tion process to be more lossy than re­cent­ly-de­ceased-brain->­plas­ti­na­tion process. It could be that warm­ing the brain up enough to plas­ti­nate does dam­age, or that the cracks caused by vit­ri­fi­ca­tion are not repara­ble and de­grade the plas­ti­nated re­sult.



  1. Threat­ens in­for­ma­tion preser­va­tion pa­ra­me­ter #2 in sev­eral ways:

    1. can plas­ti­na­tion pre­serve the level of de­tail re­quired for re­con­struc­tion? Un­known.7 The Brain Preser­va­tion Tech­nol­ogy Prize (to which I have do­nat­ed) is at­tempt­ing to spark re­search.

      Cry­on­ics as­sumes, based on anal­o­gous near-death ex­pe­ri­ences, that many things like dy­namic elec­tri­cal ac­tiv­i­ty, can be dis­re­garded for the pur­pose of per­sonal iden­tity. Plas­ti­na­tion is known to pre­serve over­all neural struc­ture in high res­o­lu­tion, as ev­i­denced by cur­rent plas­ti­na­tion tech­niques suffic­ing to cre­ate con­nec­tomes, but what does it miss? It misses the dy­namic ac­tiv­i­ty, like cry­on­ics, but cry­on­ics pre­serves things plas­ti­na­tion may not. Does plas­ti­na­tion pre­serve neu­ro­trans­mit­ter lev­els? (It seems in­con­sis­tent with the gen­eral idea of plas­ti­na­tion.) Neu­ro­trans­mit­ter lev­els change end­less­ly, but lev­els of neu­ro­trans­mit­ters can be the differ­ence be­tween san­ity and in­san­ity in the liv­ing; on the other hand, per­sonal iden­ti­ties per­sist even through ca­reers of mas­sive head trauma like box­ing or foot­ball, which affect neu­ro­trans­mit­ters (see ). What might we be miss­ing?

    2. are the meth­ods well-s­tud­ied and im­ple­mented even if they are ca­pa­ble in prin­ci­ple of pre­serv­ing the nec­es­sary in­for­ma­tion? They have been widely used in neu­ro­science, but there are no checks or ‘round trips’ show­ing that in­for­ma­tion and func­tion­al­ity is pre­served with nor­mally ex­e­cuted tech­niques - at least cry­on­ics has frozen rab­bit kid­neys to test it­self on, what does plas­ti­na­tion have?

      Coun­ter-point: brain scan­ning and the as­so­ci­ated plas­ti­na­tion tech­niques are an ex­tremely hot field of re­search, which is im­prov­ing at an amaz­ing clip akin to DNA se­quenc­ing. This ought to give us con­sid­er­able con­fi­dence in its cur­rent and fu­ture tech­niques. (This also raises an in­ter­est­ing point that any­one not dy­ing in the next decade or two is wast­ing their time by in­ves­ti­gat­ing plas­ti­na­tion. It’s en­tirely pos­si­ble that for a young or mid­dle-aged per­son, the field will ei­ther have suc­ceeded in plas­ti­nat­ing an an­i­mal or hu­man brain and then up­load­ing it, or will have dead­-ended and the fun­da­men­tal lim­its dis­cov­ered, by the time they truly need to choose be­tween cry­on­ics and plas­ti­na­tion.)

    3. Are the plas­ti­na­tion processes fast enough? Nor­mal brains are pre­served over weeks to years, which is strictly worse than a hy­po­thet­i­cal equally good process which re­quires hours. Cryo­genic cool­ing ap­pears to be in­trin­si­cally faster than chem­i­cal diffu­sion and ac­tion. How much dam­age does the ex­tra time re­quired do? (There’s some weak ev­i­dence that the rate of degra­da­tion is some­what con­stant and hence the dam­age lin­ear over time.)

  2. Threat­ens re­vival pa­ra­me­ter #5: a vit­ri­fied brain can, pre­sum­ably, be plas­ti­cized if nec­es­sary. How­ev­er, a plas­ti­cized brain is per­ma­nently plas­ti­cized. The plas­ti­cized brain has only 1 op­tion. A vit­ri­fied brain has 2 op­tions: nor­mal freez­ing and re­pair (what­ever that will be), and the plas­ti­cized route (s­can­ning and up­load, like­ly). A dis­junc­tion of two prob­a­bil­i­ties is at least as likely as ei­ther dis­junct. Ease of re­vival also affects how long stor­age must suc­ceed - if re­vival is fea­si­ble for both, but cry­on­ics is eas­ier, the cry­on­ics brain will have to last a shorter pe­riod than the plas­ti­nated brains. (This cuts both ways: if plas­ti­nated brains are eas­ier to re­vive or up­load, then it will be the cry­on­ics brains which lose some prob­a­bil­ity due to the in­creased wait­-time.)

    This may not be a large ad­van­tage for cry­on­ics. Most cry­on­ics ad­vo­cates seem to ex­pect up­load­ing will be the ul­ti­mate so­lu­tion, inas­much as brain scan­ning is ad­vanc­ing a lot faster than med­ical nan­otech­nol­ogy (see the Whole Brain Em­u­la­tion Roadmap), but there’s still a small prob­a­bil­ity that a non-u­pload or­ganic so­lu­tion will be used, and this small prob­a­bil­ity is for­feited in the plas­ti­na­tion route.


Count­ing the dis­crete items, we found 4 for plas­ti­na­tion and 2 against (yes, one point is counted twice). This is a use­less count, of course. Of those fa­vor­able 5, 2 seem to me to be prob­a­bil­ity differ­en­tials of mag­ni­tude. Of the un­fa­vor­able 3, 1 seems to be of mag­ni­tude. This count fa­vors plas­ti­na­tion as well.

I be­lieve the above fairly sets out the signs of all the re­la­tion­ships, but it is diffi­cult to fill in spe­cific num­bers for one­self, and even more diffi­cult to de­fend those num­bers.

The fun­da­men­tal ques­tion is, does the rapid ad­vance of scan­ning and the ro­bust­ness against or­ga­ni­za­tional fail­ure of plas­ti­na­tion out­weigh the risk that cry­on­ics uniquely pre­serves key in­for­ma­tion?



As­chwin de Wolf 2013:

Plas­ti­na­tion is one-way, while with proper tech­niques, the brain can be cry­on­i­cally stored such that it can later be plas­ti­nated (in case of an ex­tended emer­gency eg.); Mike Dar­win did some pre­lim­i­nary ex­per­i­ments in this area and fore­casts what such tech­niques might one day look like:

One of most diffi­cult prob­lems to be over­come when ap­ply­ing this tech­nique to a whole or­gan the size of a hu­man brain is, how do you keep the cir­cu­la­tory sys­tem ac­ces­si­ble to al­low for the re­place­ment of the wa­ter in the tis­sue with the monomer that will sub­se­quently be poly­mer­ized into a solid plas­tic, and to re­move the truly enor­mous amount of heat lib­er­ated by the exother­mic poly­mer­iza­tion re­ac­tion?

[Fig­ure 14: A cor­ro­sion cast of the cir­cu­la­tory sys­tem of the hu­man brain. The ex­ten­sive vas­cu­lar­iza­tion of the brain al­lows for use of the cir­cu­la­tory sys­tem as both a mass and heat ex­chang­er. Gas per­fu­sion of the cir­cu­la­tory sys­tem prior to cool­ing to vit­ri­fi­ca­tion tem­per­a­tures leave it ac­ces­si­ble dur­ing cryo­genic stor­age should fix­a­tion and plas­ti­na­tion be­come nec­es­sary as a fall­back po­si­tion to cry­op­reser­va­tion.]

This slide (Fig­ure 14) shows the cir­cu­la­tory sys­tem of a hu­man brain. This is the real deal, not a mod­el. What you are look­ing at is some­thing called a cor­ro­sion cast. In this case, the ar­te­r­ial cir­cu­la­tion of a hu­man brain was in­jected with a red-t­inted plas­tic ma­te­r­ial and the brain was then im­mersed in a strong base, such as a con­cen­trated so­lu­tion of sodium hy­drox­ide. The base dis­solves or cor­rodes the tis­sue away, leav­ing be­hind the red plas­tic frame­work of the ar­te­r­ial cir­cu­la­tion. Its easy to see that the hu­man brain is a strongly cir­cu­lated or­gan in fact, the brain nor­mally re­ceived 1/3rd of the rest­ing car­diac out­put about 1.5 liters of blood per minute. The FFP re­searchers de­cided that the best way to achieve both heat and mass ex­change was to keep the brains cir­cu­la­tory open and ac­ces­si­ble through­out the pro­ce­dure. In or­der to achieve this dur­ing so­lid­i­fi­ca­tion of the brain, they turned to gas per­fu­sion re­plac­ing the liq­uid in the cir­cu­la­tory sys­tem with gas.

One of the in­ves­ti­ga­tors (Mike Dar­win) re­al­ized that if the cir­cu­la­tory sys­tem of hu­man cry­on­ics pa­tients was sim­i­larly per­fused with gas dur­ing cool­ing to vit­ri­fi­ca­tion, not only would cool­ing be has­tened, thus re­duc­ing the risk of freez­ing, but the cir­cu­la­tory sys­tem of the pa­tient would re­main ac­ces­si­ble, even dur­ing stor­age at -150C. What this meant was that it would thus be the­o­ret­i­cally pos­si­ble fix and plas­ti­nate cry­on­ics pa­tients in the event that cry­op­reser­va­tion was no longer pos­si­ble.

In this sce­nar­io, a pa­tient would be re­moved from stor­age to a spe­cial ap­pa­ra­tus, the Fi­nal Fall­back Po­si­tion Sys­tem (FFPS), where his ar­te­r­ial cir­cu­la­tion would be con­nected to a re­cir­cu­lat­ing sys­tem of sol­vent chilled to -100C. This sol­vent would be pumped through the pa­tient and would be­gin dis­solv­ing the vis­cous cry­opro­tec­tan­t-wa­ter so­lu­tion in the pa­tients tis­sues. The sol­vent would also con­tain fix­a­tive ini­tially formalde­hyde to fix the pro­teins and, fi­nal­ly, a highly re­ac­tive met­al, os­mium tetrox­ide, that is nec­es­sary to fix the lipids; which com­prise both the cel­lu­lar and the in­tra­cel­lu­lar mem­branes. Once the pa­tient had been sol­vent sub­sti­tuted and fixed in this fash­ion, it would then be pos­si­ble to safely warm him up to room tem­per­a­ture and in­tro­duce the monomer re­quired for plas­ti­na­tion. In fact, if nec­es­sary, this could be done by im­mer­sion, rather by per­fu­sion (though this would ne­ces­si­tate re­moval of the brain from the head).




“Elec­tron imag­ing tech­nol­ogy for whole brain neural cir­cuit map­ping”, Hay­worth 2012:

As men­tioned above, the di­am­e­ter of neu­ronal processes rou­tinely shrink to less than 100 nm; for ex­am­ple, den­dritic spine necks and fine ax­ons can shrink down to ~40nm. With stan­dard tis­sue fix­a­tion and em­bed­ding pro­to­cols the mem­branes of these tubu­lar struc­tures are made elec­tron dense, hence to re­solve the tubu­lar na­ture of these neu­ronal processes one re­quires res­o­lu­tions on the or­der of 10 nm or less. To­day both the trans­mis­sion elec­tron mi­cro­scope (TEM) and the scan­ning elec­tron mi­cro­scope (SEM) can eas­ily achieve such res­o­lu­tion while pro­vid­ing a sig­nal to noise ra­tio suffi­cient to trace the finest neu­ropil in os­mium fixed, heavy metal stained tis­sue. How­ev­er, un­til rel­a­tively re­cently only the TEM was used for trac­ing neu­ronal cir­cuits. This re­liance on TEM has been a ma­jor road­block to at­tempts at large-s­cale au­toma­tion since TEM re­quires that sec­tions be phys­i­cally cut thin enough (<100nm) for elec­trons to pass through and that they be mounted on gos­samer thin plas­tic films through­out the imag­ing process. If scan­ning, as op­posed to trans­mis­sion, elec­tron mi­croscopy could be used it would open up many more pos­si­bil­i­ties for ro­bust au­toma­tion since then at least the thin sec­tions could be col­lected and im­aged on a thick sturdy sub­strate [Hay­worth, 2008], but be­fore the wide­spread in­tro­duc­tion of high­-bright­ness field emis­sion elec­tron sources (as op­posed to tung­sten thermionic sources) one sim­ply could not achieve SEM elec­tron probes of small enough di­am­e­ter (5 nm) and high enough cur­rent to al­low qual­ity imag­ing of neural tis­sue [Bogner et al., 2007; Joy, 1991].

In 2004, Denk and Horstman showed that this re­liance on TEM for trac­ing neural cir­cuits could in fact be over­come. They pub­lished a sem­i­nal pa­per demon­strat­ing the SBFSEM (Se­r­ial Block Face Scan­ning Elec­tron Mi­croscopy) method which could ro­bustly au­to­mate the process of ob­tain­ing a se­ries of elec­tron mi­cro­graphs from a block of neural tis­sue [Denk and Horstman, 2004]. Us­ing a high­-bright­ness field emis­sion SEM equipped with a low-en­ergy backscat­ter elec­tron de­tec­tor they first showed that one could ob­tain high­-res­o­lu­tion im­ages di­rectly from the face of the tis­sue block (thus elim­i­nat­ing the need to col­lect ul­tra­-thin sec­tion­s). Then, fol­low­ing an orig­i­nal de­sign from Leighton [1981], they built an ul­tra­mi­cro­tome into the vac­uum cham­ber of the SEM which would re­peat­edly scrape (us­ing an ex­tremely sharp di­a­mond knife) 50nm lay­ers of ma­te­r­ial of the sur­face of the tis­sue block while the SEM was used to im­age each freshly re­vealed block face at high res­o­lu­tion. The re­sult was a fully au­to­mated method to vol­ume im­age a block of neural tis­sue. They and oth­ers have con­tin­ued to re ne this tech­nique so that now it can now achieve 23nm sec­tion thick­ness and have suc­cess­fully ap­plied the tech­nique to a re­con­struc­tion of the di­rec­tion se­lec­tive cir­cuitry of the mam­malian retina [Brig­gman et al., 2011].

…In 2008, Gra­ham Knott et al. in­tro­duced the FIBSEM (Fo­cused Ion Beam Scan­ning Elec­tron Mi­croscopy) tech­nique for trac­ing neural cir­cuits [Knott et al., 2008]. This tech­nique is also known as Ion Abra­sion Scan­ning Elec­tron Mi­croscopy [Hey­mann et al., 2009]. FIBSEM works sim­i­larly to SBFSEM, but in­stead of phys­i­cally scrap­ing a thin layer of ma­te­r­ial off the block face with a di­a­mond knife, the ma­te­r­ial is in­stead ab­lated away us­ing a fo­cused beam of gal­lium ions. This change over­comes the lat­eral res­o­lu­tion lim­i­ta­tions of the SBFSEM since the FIB ab­la­tion process is rel­a­tively in­sen­si­tive to the ma­te­r­ial prop­er­ties of the em­bed­ding plas­tic and there­for a much larger elec­tron dose can be used dur­ing imag­ing. What’s more, ion beams can be tightly fo­cused achiev­ing spot sizes in the 10 nm range. Align­ing the ion beam par­al­lel to the block face so that it just grazes the sur­face al­lows re­li­able and rapid re­moval of ex­tremely thin lay­ers less than 10 nm in thick­ness. What ac­tu­ally lim­its the res­o­lu­tion of the FIBSEM (for op­ti­mally sized blocks) is the depth of pen­e­tra­tion of the imag­ing elec­trons into the sur­face of the block [Muller-Re­ichert et al., 2010] and by us­ing very low volt­ages (< 2 kV), along with sen­si­tive low-en­ergy backscat­ter elec­tron de­tec­tors with en­ergy fil­ter­ing, re­cent re­ports [Knott and Can­toni, 2011] have demon­strated depth res­o­lu­tions in the range of ~10nm with lat­eral imag­ing res­o­lu­tions of 5x5 nm. This FIBSEM voxel res­o­lu­tion of 5x5x10nm is more than suffi­cient to re­li­ably re­solve the finest neu­ronal processes and synaps­es, and should be suffi­cient to al­low for ex­tremely re­li­able au­to­mated re­con­struc­tion al­go­rithms [Luc­chi et al., 2010]. Re­cently Knott has pub­lished a Jour­nal of Vi­su­al­ized Ex­per­i­ments on­line video ar­ti­cle cov­er­ing the en­tire FIBSEM pro­ce­dure from tis­sue pro­cess­ing to au­to­mated FIBSEM milling and imag­ing, demon­strat­ing a fi­nal vol­ume im­age of a piece of cor­tex im­aged with 5x5x5 nm vox­els [Knott et al., 2011]. It is stun­ning how clearly de­mar­cated the neu­ronal processes and synapses are in the re­sult­ing FIBSEM vol­ume video.

…In sum­ma­ry, the FIBSEM tech­nique rep­re­sents a truly en­abling tech­nol­ogy for map­ping neural cir­cuits with 100% re­li­a­bil­i­ty. It can achieve voxel res­o­lu­tions of at least 5x5x10 nm, suffi­cient to al­low straight­for­ward al­go­rithms for com­puter au­to­mated trac­ing of all neu­ronal processes and iden­ti­fi­ca­tion of all synap­tic con­nec­tions [Mer­ch­n-Prez et al., 2009] along with the mor­pho­log­i­cal pa­ra­me­ters which are cor­re­lated with their strength (e.g., area of con­tact and size of post synap­tic den­si­ty). Fur­ther­more, its use of a fo­cused ion beam (as op­posed to tra­di­tional phys­i­cal sec­tion­ing with a di­a­mond knife) gives it the po­ten­tial to achieve the very high re­li­a­bil­ity lev­els needed for large-s­cale au­toma­tion. A se­ri­ous lim­i­ta­tion of to­day’s FIBSEM sys­tems is that they can achieve such high­-res­o­lu­tion im­ages only over tiny vol­umes - typ­i­cally on the or­der of a few tens of mi­crons across. How­ev­er, this lim­i­ta­tion can be rel­a­tively eas­ily over­come with the use of a loss­less sub­di­vi­sion tech­nique like the one de­scribed above. And the avail­abil­ity of such a loss­less sub­-di­vi­sion tech­nique opens up the pos­si­bil­ity of mas­sively par­al­lel FIBSEM imag­ing, some­thing that would, in any case, be nec­es­sary in or­der to im­age any rel­a­tively large vol­ume of tis­sue in a rea­son­able amount of time.

…To­day’s most typ­i­cally used tis­sue prepa­ra­tion pro­to­cols for elec­tron mi­croscopy are only able to pre­pare vol­umes of less than 1 mm3. An an­i­mal’s vas­cu­lar sys­tem is per­fused through the heart with a mix­ture of paraformalde­hyde and glu­taralde­hyde. The paraformalde­hyde quickly stops cel­lu­lar degra­da­tion and, at a slightly slower rate, the glu­taralde­hyde pro­vides stronger crosslinks to fix pro­teins in place. If per­formed care­ful­ly, such per­fu­sion through the vas­cu­lar sys­tem is able to quickly fix the en­tire ner­vous sys­tem of the an­i­mal since every cell is within a few tens of mi­crons of a cap­il­lary. How­ev­er, fol­low­ing this fix­a­tive per­fu­sion step the brain of the an­i­mal is re­moved and a very small piece dis­sected to un­dergo the re­main­ing tis­sue prepa­ra­tion steps - which are typ­i­cally all per­formed by sim­ple im­mer­sion in chem­i­cals. These steps in­clude im­mer­sion in os­mium tetrox­ide (to fix mem­brane lipid mol­e­cules in place), im­mer­sion in heavy metal stain­ing so­lu­tions (e.g., uranyl ac­etate), im­mer­sion in a graded se­ries of al­co­hols (to re­move wa­ter from the tis­sue), and fi­nally im­mer­sion in a plas­tic resin dis­solved in an or­ganic sol­vent (to com­pletely in­fil­trate the tis­sue with the heat-cur­able resin). Be­cause these steps are per­formed by sim­ple im­mer­sion (and thus diffu­sion) the process fails if at­tempted on blocks larger than 1 mm3. The re­sult is de­struc­tion of tis­sue ul­tra­struc­ture and poor stain­ing in the depths of the block.

Ob­vi­ously such vol­ume lim­i­ta­tions must be over­come be­fore hu­man mind up­load­ing can be at­tempt­ed. Al­though it has never been demon­strated that a whole mam­malian brain can be pre­served at the ul­tra­struc­ture level for elec­tron mi­cro­scopic imag­ing, there are many re­sults that sug­gest that a pro­to­col could be de­vel­oped to do just that. Per­fu­sion fix­a­tion with os­mium tetrox­ide has been demon­strated on whole brains [Palay et al., 1962], and se­r­ial vas­cu­lar per­fu­sion first with glu­taralde­hy­de, fol­lowed by os­mium tetrox­ide and uranyl ac­etate, and fi­nally by an al­co­hol de­hy­dra­tion se­ries has been demon­strated on whole or­gans [Ba­chofen et al., 1982; Old­mixon et al., 1985] demon­strat­ing suffi­cient preser­va­tion to al­low ul­tra­struc­ture stud­ies. Plas­tic in­fil­tra­tion of whole mouse brains has also been demon­strated [May­erich et al., 2008], and there are re­cent re­ports of the de­vel­op­ment of a full ul­tra­struc­ture fix­a­tion, stain­ing, and em­bed­ding pro­to­col for the mouse brain for use in the map­ping of long dis­tance axon tra­jec­to­ries via se­r­ial block face SEM [Mikula et al., 2011]. There is even a chal­lenge prize be­ing offered for the first demon­stra­tion of such ul­tra­struc­ture preser­va­tion across an en­tire large mam­malian brain [Hay­worth, 2011]. Given these re­sults and the fact that the bur­geon­ing field of Con­nec­tomics will re­quire larger and larger vol­umes of high­-qual­ity pre­pared tis­sue, it is likely that a pro­to­col to pre­serve and plas­tic-em­bed an en­tire hu­man brain at the ul­tra­struc­ture level will be per­fected rel­a­tively soon.


  1. As far as I can tell, pretty much every pro-con for plas­ti­na­tion vs cry­on­ics ap­plies to chem­i­cal fix­a­tion with the ex­cep­tion of the lipids, so in the fol­low­ing I treat them as syn­ony­mous.↩︎

  2. Ap­par­ently it’s best to store even a plas­ti­cized brain in cryo­genic stor­age; Jor­dan Sparks says “If they don’t tran­si­tion to cry­op­reser­va­tion, dam­age will be on­go­ing for decades.”↩︎

  3. “The Ar­mories of the Lat­ter Day La­putas, Part 5” has the sta­tis­tics. Be­sides 98% of star­tups dy­ing, es­tab­lished com­pa­nies die fre­quent­ly: “…if we re-set the graph at 5 years, and then fol­low the re­main­ing co­hort of en­ter­prises out to the 10 year mark, the mor­tal­ity rate is still quite high with only 29% of busi­nesses sur­viv­ing.” The to­tal mor­tal­ity is con­sid­er­able: “Thus, the chances of a busi­ness en­tity (ex­clud­ing re­li­gious and aca­d­e­mic in­sti­tu­tions) sur­viv­ing for >100 years is 1.096%” One some­times see peo­ple pro­vide their own val­ues for a cry­on­ics Drake equa­tion - often the re­sult is a com­fort­ing 1-5%. This shows they ex­pect to be re­vived very soon, think cry­on­ics or­ga­ni­za­tions are ex­empt from these sta­tis­tics, or are un­aware. Be­ing non-profit helps only a lit­tle: “How­ev­er, by the 30 year mark, ~95% of NPOs have failed.”↩︎

  4. The brain of the , processed by The Brain Ob­ser­va­tory of UC San Diego.↩︎

  5. Al­though hope­fully in ways more effi­cient than a few pro­fes­sors and tech­ni­cians la­bor­ing to­gether or yok­ing dozens of stu­dents! In this vein, the Mouse Brain Ar­chi­tec­ture Project is some­thing to watch - what can be done to a mouse brain may, a few gen­er­a­tions lat­er, be done to a hu­man brain.↩︎

  6. I have a gen­eral skep­ti­cism about ap­plied med­i­cine and bi­ol­ogy (and re­viv­i­fi­ca­tion is very ap­plied), from a life­time of bro­ken promises and failed pre­dic­tions about the com­ing fruits of med­i­cine and bi­ol­o­gy. I re­cently ran across an ex­am­ple in a by Don­ald Brown, then an ac­com­plished pro­fes­sor of bi­ol­ogy at Johns Hop­kins:

    We don’t know yet how these genes work, but tech­niques of mod­ern ge­net­ics which led to the dis­cov­ery of these genes in the first place will pro­vide these in­sights. In the next ten years we are go­ing to un­der­stand bet­ter and per­haps even cure some of the most se­ri­ous dis­eases that afflict mankind, such as di­a­betes, ar­te­rioscle­ro­sis, par­a­sitic dis­eases, the com­mon cold, cys­tic fi­bro­sis, cer­tain kinds of arthri­tis, im­mune dis­eases, and in­fec­tious dis­eases, just to name a few. A mol­e­c­u­lar ba­sis for at least some kinds of schiz­o­phre­nia will be found. We will learn about the bio­chem­istry of the ag­ing process, which also has a strong ge­netic com­po­nent. This does­n’t guar­an­tee pro­lon­ga­tion of life, but rather an im­prove­ment of the qual­ity of life in old age. We need sen­si­tive as­says for the effects of chem­i­cals, pol­lu­tants, and drugs as causative agents of birth de­fects like those de­vel­oped to de­ter­mine car­cino­genic po­ten­tial. There have yet to be de­vel­oped sim­ple, safe, and re­versible con­tra­cep­tives for males.

    This reads like it was writ­ten yes­ter­day, and not in 1984 - more than 27 years ago. The pre­dic­tions about the Hu­man Genome Project have fared lit­tle bet­ter. My rule of thumb is that in the fu­ture, our un­der­stand­ing and in­for­ma­tion will out­strip our lab­o­ra­tory pro­to­types by even more than to­day, and the pro­to­types as much out­strip the gen­er­ally avail­able prod­ucts; hence, I am un­sur­prised by the as­tound­ing progress in DNA se­quenc­ing, and equally un­sur­prised by the as­tound­ing dearth of new drugs and treat­ments. For­tu­nately for plas­ti­nated brains, enough in­for­ma­tion and un­der­stand­ing can make up for sta­sis in other ar­eas; as long as com­put­ers and scan­ning tech­nol­ogy con­tinue to ad­vance, things may yet work out for them.↩︎

  7. From Per­ry’s ar­ti­cle about chem­i­cal fix­a­tion:

    As with cry­on­ics it­self, the ba­sic an­swers are un­known. Some en­cour­age­ment is pro­vided by the high level of de­tail seen in pre­served brain sam­ples us­ing, for ex­am­ple, formalde­hyde fix­a­tion. Ul­tra­struc­tural de­tails un­der the high mag­ni­fi­ca­tion of elec­tron mi­croscopy (10,000x plus) are quite clear, though this alone is not a demon­stra­tion that all the de­tails one would like are pre­sent. How­ev­er, the same prob­lem ex­ists with tis­sue pre­served cryo­geni­cally the an­swer to whether the preser­va­tion cap­tures fine enough de­tails is un­known though there are at least some en­cour­ag­ing signs along with rea­sons for con­cern.

    He is not so san­guine about plas­ti­na­tion;

    A pos­si­ble draw­back of this ap­proach, from the stand­point of pre­serv­ing the fine struc­tures that are es­pe­cially im­por­tant from a cry­on­ics stand­point, is the rel­a­tively harsh reg­i­men needed to pro­duce the fin­ished prod­uct. Typ­i­cal­ly, the process starts with an alde­hy­de-fixed spec­i­men in aque­ous so­lu­tion. The spec­i­men is placed in ace­tone, and suc­ces­sive changes of the bath re­move wa­ter and fats. Fi­nally the resin monomer is in­tro­duced, the re­main­ing ace­tone is re­moved by vac­u­um, and in­duced catal­y­sis yields the de­sired poly­mer­iza­tion. Con­cerns have been raised about whether de­fat­ting would oblit­er­ate im­por­tant brain in­for­ma­tion, though there does not ap­pear to be strong ev­i­dence of this. (Here it is ap­pro­pri­ate to men­tion that lipids nev­er­the­less could con­tain im­por­tant in­for­ma­tion; preser­va­tion of lipids is a diffi­cult process that has not been cov­ered in this pre­lim­i­nary sur­vey but de­serves con­sid­er­a­tion.)

    Ol­son men­tions lipids in pass­ing:

    A sec­ond ex­am­ple of re­dun­dancy in­volves the lo­ca­tions of the neu­ronal mem­branes (i.e. the neu­ronal con­fig­u­ra­tion). The in­for­ma­tion of the mem­branes’ po­si­tions is con­tained not only in the phys­i­cal po­si­tions of the mem­brane lipids, but also in the cel­lu­lar cy­toskele­ton (which is made of pro­teins) whose pur­pose is, among other things, to hold the mem­brane in its con­fig­u­ra­tion (15). Thus, even if a sub­stan­tial pro­por­tion of lipids were ex­tracted in the course of chem­i­cal preser­va­tion of a brain (as is the case with some preser­va­tion tech­niques), it is plau­si­ble that the in­for­ma­tion of the neu­ronal con­fig­u­ra­tion would still ex­ist in the crosslinked cy­toskele­tons of the neu­rons.