adapting registration for 2D time series

[kellyhill90]

I am trying to adapt the registration for a 2d time series. I started from the BlockRegistrationScheduler stack-by-stack example and tried to identify all the places that 3d data needs to be converted to 2d. However, I still end up with an error that seems to be related to size mismatch. Any suggestions? Error and code both included below. (Probably not relevant, but note that I skipped steps related to preprocessing, bias adjustment, and roi selection.)

ERROR: ssize and gridsize must have the same length
aperture_grid(::Tuple{Int64,Int64,Int64}, ::Tuple{Int64,Int64}) at /home/kelly/.julia/v0.6/BlockRegistration/src/RegisterMismatchCommon.jl:153
worker(::RegisterWorkerApertures.Apertures{Array{ColorTypes.Gray{FixedPointNumbers.Normed{UInt16,16}},2},Float64,StepRangeLen{Float64,Base.TwicePrecision{Float64},Base.TwicePrecision{Float64}},2}, ::Array{ColorTypes.Gray{FixedPointNumbers.Normed{UInt16,16}},3}, ::Int64, ::Dict{Symbol,Any}) at /home/kelly/.julia/v0.6/BlockRegistrationScheduler/src/RegisterWorkerApertures.jl:163
macro expansion at /home/kelly/.julia/v0.6/BlockRegistrationScheduler/src/RegisterDriver.jl:90 [inlined]
(::RegisterDriver.##3#9{JLD.JldFile,Dict{Symbol,Any},SharedArray{Bool,1},SharedArray{Bool,1},RegisterDriver.#getnextidx#8,RemoteChannel{Channel{Any}},Array{RegisterWorkerApertures.Apertures,1},Array{ColorTypes.Gray{FixedPointNumbers.Normed{UInt16,16}},3},Array{Dict{Symbol,Any},1}})() at ./task.jl:335
Stacktrace:
 [1] sync_end() at ./task.jl:287
 [2] macro expansion at ./task.jl:303 [inlined]
 [3] (::RegisterDriver.##2#7{Array{RegisterWorkerApertures.Apertures,1},Array{ColorTypes.Gray{FixedPointNumbers.Normed{UInt16,16}},3},Array{Dict{Symbol,Any},1},Int64})(::JLD.JldFile) at /home/kelly/.julia/v0.6/BlockRegistrationScheduler/src/RegisterDriver.jl:78
 [4] #jldopen#11(::Array{Any,1}, ::Function, ::RegisterDriver.##2#7{Array{RegisterWorkerApertures.Apertures,1},Array{ColorTypes.Gray{FixedPointNumbers.Normed{UInt16,16}},3},Array{Dict{Symbol,Any},1},Int64}, ::String, ::Vararg{String,N} where N) at /home/kelly/.julia/v0.6/JLD/src/JLD.jl:256
 [5] driver(::String, ::Array{RegisterWorkerApertures.Apertures,1}, ::Array{ColorTypes.Gray{FixedPointNumbers.Normed{UInt16,16}},3}, ::Array{Dict{Symbol,Any},1}) at /home/kelly/.julia/v0.6/BlockRegistrationScheduler/src/RegisterDriver.jl:68

For reference, this is how I adapted the code. The ssize/gridsize error comes at the last line (driver).

using BlockRegistration, BlockRegistrationScheduler
using RegisterWorkerApertures

const μm = u"μm"
const s=u"s"

img0=load("testM.tif");
fixedidx = (size(img0,3)+1) >> 1
fixed0=img0[:,:,fixedidx];
ps=[1.15μm,1.15μm]
σ = 25μm
sigmahp = Float64[σ/x for x in ps]
sigmalp = [0,0]  # lowpass filtering is not currently recommended
fixed=fixed0;

gridsize=(round(Int,Float64[10μm/x for x in ps])[1],round(Int,Float64[10μm/x for x in ps])[2])
knots = map(d->linspace(1,size(fixed,d),gridsize[d]), (1:ndims(fixed)...))

mxshift=(10,20);
λ=0.01

algorithm = Apertures[Apertures(fixed, knots, mxshift, λ, correctbias=false)]

mon = monitor(algorithm, (), Dict{Symbol,Any}(:u=>ArrayDecl(Array{Vec{2,Float64},2}, gridsize)))

basename = splitext(splitdir(fn)[2])[1]
@show basename
fileout = string(basename, ".register")

@time driver(fileout, algorithm, img0, mon)

[mdhe1248]

Kelly, Could you add Axis on your image? Something like this:

fn = "testM.tif"
img0=load(fn);
img = AxisArray(img0, (:x, :y, :time), (1.15μm, 1.15μm, 2s))

I think may help solving the error above. However, you may get another error.

[mdhe1248]

This is the error I get:

julia> @time driver(fileout, algorithm, img, mon)
Working on 1
Planning FFTs (maximum Inf seconds)...done (0.99 seconds)
ERROR: MethodError: Cannot `convert` an object of type SVector{2,Float64} to an object of type Tuple{Float64,Float64}
This may have arisen from a call to the constructor Tuple{Float64,Float64}(...),
since type constructors fall back to convert methods.
copy!(::IndexLinear, ::Array{FixedSizeArrays.Vec{2,Float64},2}, ::IndexLinear, ::Array{SVector{2,Float64},2}) at ./abstractarray.jl:655
copy!(::SharedArray{FixedSizeArrays.Vec{2,Float64},2}, ::Array{SVector{2,Float64},2}) at ./sharedarray.jl:558
monitor!(::Dict{Symbol,Any}, ::Symbol, ::Array{SVector{2,Float64},2}) at /home/donghoon/.julia/v0.6/BlockRegistrationScheduler/src/RegisterWo
rkerShell.jl:105
worker(::RegisterWorkerApertures.Apertures{AxisArrays.AxisArray{ColorTypes.Gray{FixedPointNumbers.Normed{UInt16,16}},2,SubArray{ColorTypes.Gr
ay{FixedPointNumbers.Normed{UInt16,16}},2,Array{ColorTypes.Gray{FixedPointNumbers.Normed{UInt16,16}},3},Tuple{Base.Slice{Base.OneTo{Int64}},B
ase.Slice{Base.OneTo{Int64}},Int64},true},Tuple{AxisArrays.Axis{:x,StepRangeLen{Quantity{Float64, Dimensions:{𝐋}, Units:{μm}},Quantity{Float6
4, Dimensions:{𝐋}, Units:{μm}},Quantity{Float64, Dimensions:{𝐋}, Units:{μm}}}},AxisArrays.Axis{:y,StepRangeLen{Quantity{Float64, Dimensions:{
𝐋}, Units:{μm}},Quantity{Float64, Dimensions:{𝐋}, Units:{μm}},Quantity{Float64, Dimensions:{𝐋}, Units:{μm}}}}}},Float64,StepRangeLen{Float64,
Base.TwicePrecision{Float64},Base.TwicePrecision{Float64}},2}, ::AxisArrays.AxisArray{ColorTypes.Gray{FixedPointNumbers.Normed{UInt16,16}},3,
Array{ColorTypes.Gray{FixedPointNumbers.Normed{UInt16,16}},3},Tuple{AxisArrays.Axis{:x,StepRangeLen{Quantity{Float64, Dimensions:{𝐋}, Units:{
μm}},Quantity{Float64, Dimensions:{𝐋}, Units:{μm}},Quantity{Float64, Dimensions:{𝐋}, Units:{μm}}}},AxisArrays.Axis{:y,StepRangeLen{Quantity{F
loat64, Dimensions:{𝐋}, Units:{μm}},Quantity{Float64, Dimensions:{𝐋}, Units:{μm}},Quantity{Float64, Dimensions:{𝐋}, Units:{μm}}}},AxisArrays.
Axis{:time,StepRange{Quantity{Int64, Dimensions:{𝐓}, Units:{s}},Quantity{Int64, Dimensions:{𝐓},Units:{s}}}}}}, ::Int64, ::Dict{Symbol,Any}) a
t /home/donghoon/.julia/v0.6/BlockRegistrationScheduler/src/RegisterWorkerApertures.jl:190
macro expansion at /home/donghoon/.julia/v0.6/BlockRegistrationScheduler/src/RegisterDriver.jl:90 [inlined]
(::RegisterDriver.##3#9{JLD.JldFile,Dict{Symbol,Any},SharedArray{Bool,1},SharedArray{Bool,1},RegisterDriver.#getnextidx#8,RemoteChannel{Chann
el{Any}},Array{RegisterWorkerApertures.Apertures,1},AxisArrays.AxisArray{ColorTypes.Gray{FixedPointNumbers.Normed{UInt16,16}},3,Array{ColorTy
pes.Gray{FixedPointNumbers.Normed{UInt16,16}},3},Tuple{AxisArrays.Axis{:x,StepRangeLen{Quantity{Float64, Dimensions:{𝐋}, Units:{μm}},Quantity
{Float64, Dimensions:{𝐋}, Units:{μm}},Quantity{Float64, Dimensions:{𝐋}, Units:{μm}}}},AxisArrays.Axis{:y,StepRangeLen{Quantity{Float64, Dimen
sions:{𝐋}, Units:{μm}},Quantity{Float64, Dimensions:{𝐋}, Units:{μm}},Quantity{Float64, Dimensions:{𝐋}, Units:{μm}}}},AxisArrays.Axis{:time,St
epRange{Quantity{Int64, Dimensions:{𝐓}, Units:{s}},Quantity{Int64, Dimensions:{𝐓}, Units:{s}}}}}},Array{Dict{Symbol,Any},1}})() at ./task.jl:
335
Stacktrace:
 [1] sync_end() at ./task.jl:287
 [2] macro expansion at ./task.jl:303 [inlined]
 [3] (::RegisterDriver.##2#7{Array{RegisterWorkerApertures.Apertures,1},AxisArrays.AxisArray{ColorTypes.Gray{FixedPointNumbers.Normed{UInt16,
16}},3,Array{ColorTypes.Gray{FixedPointNumbers.Normed{UInt16,16}},3},Tuple{AxisArrays.Axis{:x,StepRangeLen{Quantity{Float64, Dimensions:{𝐋}, 
Units:{μm}},Quantity{Float64, Dimensions:{𝐋}, Units:{μm}},Quantity{Float64, Dimensions:{𝐋}, Units:{μm}}}},AxisArrays.Axis{:y,StepRangeLen{Qua
ntity{Float64, Dimensions:{𝐋}, Units:{μm}},Quantity{Float64, Dimensions:{𝐋}, Units:{μm}},Quantity{Float64, Dimensions:{𝐋}, Units:{μm}}}},Axis
Arrays.Axis{:time,StepRange{Quantity{Int64, Dimensions:{𝐓}, Units:{s}},Quantity{Int64, Dimensions:{𝐓}, Units:{s}}}}}},Array{Dict{Symbol,Any},
1},Int64})(::JLD.JldFile) at /home/donghoon/.julia/v0.6/BlockRegistrationScheduler/src/RegisterDriver.jl:78
 [4] #jldopen#11(::Array{Any,1}, ::Function, ::RegisterDriver.##2#7{Array{RegisterWorkerApertures.Apertures,1},AxisArrays.AxisArray{ColorType
s.Gray{FixedPointNumbers.Normed{UInt16,16}},3,Array{ColorTypes.Gray{FixedPointNumbers.Normed{UInt16,16}},3},Tuple{AxisArrays.Axis{:x,StepRang
eLen{Quantity{Float64, Dimensions:{𝐋}, Units:{μm}},Quantity{Float64, Dimensions:{𝐋}, Units:{μm}},Quantity{Float64, Dimensions:{𝐋}, Units:{μm}
}}},AxisArrays.Axis{:y,StepRangeLen{Quantity{Float64, Dimensions:{𝐋}, Units:{μm}},Quantity{Float64, Dimensions:{𝐋}, Units:{μm}},Quantity{Floa
t64, Dimensions:{𝐋}, Units:{μm}}}},AxisArrays.Axis{:time,StepRange{Quantity{Int64, Dimensions:{𝐓}, Units:{s}},Quantity{Int64, Dimensions:{𝐓},
 Units:{s}}}}}},Array{Dict{Symbol,Any},1},Int64}, ::String, ::Vararg{String,N} where N) at /home/donghoon/.julia/v0.6/JLD/src/JLD.jl:256
 [5] driver(::String, ::Array{RegisterWorkerApertures.Apertures,1}, ::AxisArrays.AxisArray{ColorTypes.Gray{FixedPointNumbers.Normed{UInt16,16
}},3,Array{ColorTypes.Gray{FixedPointNumbers.Normed{UInt16,16}},3},Tuple{AxisArrays.Axis{:x,StepRangeLen{Quantity{Float64, Dimensions:{𝐋}, Un
its:{μm}},Quantity{Float64, Dimensions:{𝐋}, Units:{μm}},Quantity{Float64, Dimensions:{𝐋}, Units:{μm}}}},AxisArrays.Axis{:y,StepRangeLen{Quant
ity{Float64, Dimensions:{𝐋}, Units:{μm}},Quantity{Float64, Dimensions:{𝐋}, Units:{μm}},Quantity{Float64, Dimensions:{𝐋}, Units:{μm}}}},AxisAr
rays.Axis{:time,StepRange{Quantity{Int64, Dimensions:{𝐓}, Units:{s}},Quantity{Int64, Dimensions:{𝐓}, Units:{s}}}}}}, ::Array{Dict{Symbol,Any}
,1}) at /home/donghoon/.julia/v0.6/BlockRegistrationScheduler/src/RegisterDriver.jl:68

Does anyone have an idea?

[kellyhill90]

Jerry thanks for the idea. I get the same new error that you posted and haven't been able to work through a solution to it either. We'll see if anyone has another thought.

[timholy]

Fixes coming!

[kellyhill90]

I'm trying to write a header file to open the warped output from a 2D registration. My original data was a tif file, not Imagine format. My best guess with some help from Jerry and Cody is:

props=imgoriginal.properties
props =Dict("headeronly" => true, "datafile" => imgwarped)
save("test.nhdr", imgoriginal; props = props)  #imgoriginal is just used a placeholder here; imgwarped will be loaded based on the headerfile

However the first line is unsuccessful; either ERROR: "type Array has no field properties" or ERROR: "type AxisArray has no field properties" (depending whether I loaded the tif file directly or converted it to AxisArray). If I skip the first line completely and just set the two properties (headeronly, datafile) explicitly, then the resulting header file loads as a time series of random noise.

Is there another solution to reach the end goal of opening the warped image? How to load properties of an AxisArray if not by img.properties?

[mdhe1248]

Okay. Tim taught me how to rewrite a header before. I found how I did it. I did not fully test with Julia06 yet. I am working on it. It will be quick.

datafn = "warp_pre.cam"  ### Warped image file name.
save_headerfn = "warp_pre.nhdr"    #### new header name.

####if you have img with AxisArray, `ax = img.axes` may work instead of the
steps below.
psx = (1.2:1.2:512*1.2)*μm
psy = (1.2:1.2:512*1.2)*μm
ti = (2:2:16*2)*s
ax = (Axis{:x}(psx), Axis{:y}(psy), Axis{:time}(ti))

header = NRRD.headerinfo(Float16, ax)  ## Maybe in your case,` Float32`
header["datafile"] = datafn

#### Save header
open(save_headerfn, "w") do file
    write(file, magic(format"NRRD")) # Because header does not know NRRD
format.
    NRRD.write_header(file, "0004", header)  #"0004" is newer version.
"0002" is older version.
end

UPDATE:: It works on my Cannon environment. (with subtle modifications e.g. filenames, using modules)

2017-07-06 10:07 GMT-05:00 kellyhill90 notifications@github.com:

I'm trying to write a header file to open the warped output from a 2D registration. My original data was a tif file, not Imagine format. My best guess with some help from Jerry and Cody is:

props=imgoriginal.properties props =Dict("headeronly" => true, "datafile" => imgwarped) save("test.nhdr", imgoriginal; props = props) #imgoriginal is just used a placeholder here; imgwarped will be loaded based on the headerfile

However the first line is unsuccessful; either ERROR: "type Array has no field properties" or ERROR: "type AxisArray has no field properties" (depending whether I loaded the tif file directly or converted it to AxisArray). If I skip the first line completely and just set the two properties (headeronly, datafile) explicitly, then the resulting header file loads as a time series of random noise.

Is there another solution to reach the end goal of opening the warped image? How to load properties of an AxisArray if not by img.properties?

— You are receiving this because you commented. Reply to this email directly, view it on GitHub https://github.com/HolyLab/BlockRegistrationScheduler/issues/38#issuecomment-313424410, or mute the thread https://github.com/notifications/unsubscribe-auth/AMvfFWtJEHINA_gUBENKFzf3xDj82W_gks5sLPgPgaJpZM4OJ9dw .

-- Donghoon (Jerry) Lee Neuroscience PhD Candidate | Lab of Timothy E. Holy Department of Anatomy and Neurobiology Graduate Program in Neuroscience Washington University in St. Louis

[kellyhill90]

What packages do you use for this to work? Do I need to clone this NRRD package directly? https://github.com/JuliaIO/NRRD.jl

using Images, FileIO --> I get UndefVarError: NRRD not defined

Maybe there is another package that I need to add, clone, or update before using NRRD?

[mdhe1248]

Pkg.add("NRRD") using NRRD

May solve the issue.

On Jul 6, 2017 1:36 PM, "kellyhill90" notifications@github.com wrote:

Do you think we need to clone this NRRD package directly? https://github.com/JuliaIO/NRRD.jl

Even after Using Images, FileIO, I still get UndefVarError: NRRD not defined

— You are receiving this because you commented.

Reply to this email directly, view it on GitHub https://github.com/HolyLab/BlockRegistrationScheduler/issues/38#issuecomment-313482416, or mute the thread https://github.com/notifications/unsubscribe-auth/AMvfFQB-T-xe9uGedAqx8C4MYjiAgrFDks5sLSkpgaJpZM4OJ9dw .

[kellyhill90]

Fantastic. Got it working. Now I can play around with optimization. Thanks for your help, Jerry!

[timholy]

@mdhe1248, can I copy your example and paste to the NRRD README? Seems worth documenting somewhere (and I've been lazy about documenting it).

[mdhe1248]

Sure!

On Jul 7, 2017 5:26 PM, "Tim Holy" notifications@github.com wrote:

@mdhe1248 https://github.com/mdhe1248, can I copy your example and paste to the NRRD README? Seems worth documenting somewhere (and I've been lazy about documenting it).

— You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub https://github.com/HolyLab/BlockRegistrationScheduler/issues/38#issuecomment-313807359, or mute the thread https://github.com/notifications/unsubscribe-auth/AMvfFfbrwtAE0P0yebEk4xmqScU4YCOvks5sLrCKgaJpZM4OJ9dw .