How to do iterative reconstruction after the projection splicing obtained by modular projection beam？

Nobody has claimed this yet.

Hi Kyle,

I have encountered some issues with iterative reconstruction of projections created by the `modularbeam` function recently.

I am trying to simulate a static CT imaging geometry with a fixed arc-shaped detector and a fixed arc-shaped ray source. Referring to your demo_9, I set up each module with only one detector unit, and by combining multiple modules, I formed the arc-shaped detector. I defined the coordinates for each module of the detector and ensured that each module corresponds to the same ray source position. By using the `set_modularbeam` function and forward projection function, I obtain the projection for a single ray source position, and then by looping through all ray source positions, I get the projections for each ray source position. Finally, I stitch them together to get the final projection.

However, there is an issue: each time the loop runs, the `set_modularbeam` function saves the imaging geometry for only a single ray source position, which does not correspond to the complete imaging geometry for the final projection. I am unsure how to directly apply the iterative algorithm for reconstruction on the final projection.

I hope to receive some guidance from you. Thank you!

```

moduleCenters = np.zeros((numSticks,3),dtype=np.float32)
colVectors = np.zeros((numSticks,3), dtype=np.float32)
rowVectors = np.zeros((numSticks,3), dtype=np.float32)
rowVectors[:,2] = 1
for m in range(numSticks):
colVectors[m,0] = -np.cos(alphas[m])
colVectors[m,1] = np.sin(alphas[m])
moduleCenters[m,0] = odd*np.sin(alphas[m])
moduleCenters[m,1] = -odd*np.cos(alphas[m])
for n in range(numAngles): ##Different angles correspond to different ray source positions
phi = phis[n]*np.pi/180-np.pi/2
sourcePositions = np.zeros((numSticks,3),dtype=np.float32)
sourcePositions[:,0] = sod*np.sin(phi)
sourcePositions[:,1] = sod*np.cos(phi)
sourcePositions[:,2] = 0
leapct_true.set_modularbeam(numSticks, numRows, numCols_per_module, pixelSize, pixelSize, sourcePositions, moduleCenters, rowVectors, colVectors)
g_sim = leapct_true.allocateProjections() # shape is numAngles, numRows, numCols
# "Simulate" the data
leapct_true.project(g_sim,phantom)
# Concatenate the projections onto each modular into one projection
for m in range(numSticks):
g[n,:,m*numCols_per_module:(m+1)*numCols_per_module] = g_sim[m,:,:]

```

Start with demo_9 and the documented set_modularbeam, allocateProjections, and project entry points. Compare the single-source-position loop with the stitched projection layout, then determine the supported iterative-reconstruction workflow for this modular geometry. Done means establishing a reproducible way to reconstruct the final projection or clearly documenting the limitation.

Written by the indexing model from the issue text.