--- kernelspec: display_name: IDL [conda env:gdl] * language: IDL name: conda-env-gdl-idl --- ## GDL demo notebook +++ Demonstration of GDL [(gnudatalanguage)](https://github.com/gnudatalanguage/gdl) This notebook creates a Shepp-Logan phantom, projects it and then performs an FBP reconstruction. ```{code-cell} ;; L.A. Shepp and B.F. Logan, “The Fourier reconstruction of a head section,” ;; IEEE Trans. Nucl. Sci. 21(3), 21–43 (1974). function shepplogan, size = size if NOT keyword_set(size) then size = 256 phantom = fltarr(size,size) tmp = (findgen(size)-((size-1)/2.0)) / (size/2.0) xcor = rebin(tmp,size,size) ycor = rebin(transpose(tmp),size,size) tmp = fltarr(size,size) aa={cx: 0.0, cy: 0.0, maj:0.69, min:0.92, theta: 0.0, val: 2.0 } bb={cx: 0.0, cy:-0.0184, maj:0.6624, min:0.874, theta: 0.0, val:-0.98} cc={cx: 0.22, cy: 0.0, maj:0.11, min:0.31, theta:-18.0, val:-0.02} dd={cx:-0.22, cy: 0.0, maj:0.16, min:0.41, theta: 18.0, val:-0.02} ee={cx: 0.0, cy: 0.35, maj:0.21, min:0.25, theta: 0.0, val:-0.01} ff={cx: 0.0, cy: 0.1, maj:0.046, min:0.046, theta: 0.0, val:-0.01} gg={cx: 0.0, cy:-0.1, maj:0.046, min:0.046, theta: 0.0, val:-0.01} hh={cx:-0.08, cy:-0.605, maj:0.046, min:0.023, theta: 0.0, val:-0.01} ii={cx: 0.0, cy:-0.605, maj:0.023, min:0.023, theta: 0.0, val:-0.01} jj={cx: 0.06, cy:-0.605, maj:0.023, min:0.046, theta: 0.0, val:-0.01} list = [aa,bb,cc,dd,ee,ff,gg,hh,ii,jj] for n = 0, n_elements(list)-1 do begin tmp = ((xcor-list[n].cx) / list[n].maj)^2 $ + ((ycor-list[n].cy) / list[n].min)^2 if list[n].theta NE 0 then begin nx = (size-1) * (list[n].cx + 1) / 2 ny = (size-1) * (list[n].cy + 1) / 2 tmp = rot(tmp, -list[n].theta, 1, nx, ny, /interp, /pivot) endif phantom[where(tmp LE 1.0)] += list[n].val endfor return, phantom < 1.1 end ``` ```{code-cell} ;; Enable inline plotting !inline=1 ``` ```{code-cell} phantom = shepplogan() window, 0, xsize=256, ysize=256 tvscl, phantom > 0.95 ``` Now we define simple forward and backprojection functions: ```{code-cell} function proj, image, angle sino_size_x = max(size(image,/dim)) sino_size_y = n_elements(angle) sino = fltarr(sino_size_x, sino_size_y) for aa=0, n_elements(angle)-1 do begin sino[*,aa] = total(rot(image, angle[aa], /interp), 2) endfor return, sino end ``` ```{code-cell} function back, sino, angle image_size = n_elements(sino[*,0]) image = fltarr(image_size,image_size) for aa=0, n_elements(angle)-1 do begin image += rot(rebin(sino[*,aa],[image_size,image_size]), -angle[aa], /interp) endfor return, image / n_elements(angle) end ``` This allows us to create a sinogram: ```{code-cell} angles = findgen(360) sino = proj(phantom, angles) window, 0, xsize=256, ysize=360 tvscl, sino ``` On this we can apply a simple FBP reconstruction: ```{code-cell} ;; G.L. Zeng, “Revisit of the Ramp Filter,” ;; IEEE Trans. Nucl. Sci. 62(1), 131–136 (2015). function fbp, sino, angles ntot = n_elements(sino[*,0]) nang = n_elements(sino[0,*]) npos = ntot / 2 + 1 ; integer division needed ! nneg = ntot - npos freq = findgen(ntot) freq[npos:ntot-1] = REVERSE(freq[1:nneg]) freq[0] = 1 filter = -1 / (!pi * freq)^2 filter[where(freq mod 2 EQ 0)] *= -0.0 filter[0] = 0.25 filter = abs(fft(filter)) * ntot filter[ntot/4:ntot/4+ntot/2-1] *= 0.0 filter = rebin(filter,ntot,nang) ;; apply filter to the sinogram fsino = fft(sino, dim=1) fsino *= filter fsino = fft(fsino, dim=1, /overwrite, /inverse) fsino = !pi * real_part(fsino) ;; backproject the filtered sinogram return, back(fsino, angles) end ``` ```{code-cell} reconstruction = fbp(sino, angles) window, 0, xsize=512, ysize=256 tvscl, [phantom, reconstruction] > 0.8 ``` ### The end