Forward projection bug related to the voxel spacing

[oooo1114]

Hi, I have tested LEAP with my geometry. When the voxel spacing (width or height) is smaller than the real pixel width or height in the isocenter, i.e. pixel width or height * sod / sdd, the forward projection goes wrong. Most views are all zeroes, except for tens of weird views near the center of angles. Here are the scripts which can reproduce the bug:

import sys
import os
import time
import numpy as np
from leapctype import *
leapct = tomographicModels()
numCols = 860
numTurns = 6
numAngles = 8640
numRows = 32
pixelHeight = 1.12
pixelWidth = 1.0645
centerRow = 0.5*(numRows-1)
centerCol = 0.5*(numCols-1)-4.2412
phis = leapct.setAngleArray(numAngles, 360.0*numTurns)
sod = 570
sdd = 1005
pitch = 0.861
leapct.set_conebeam(numAngles, numRows, numCols, pixelHeight, pixelWidth, centerRow, centerCol, phis, sod, sdd)
leapct.set_normalizedHelicalPitch(pitch)
leapct.set_curvedDetector()

numX = 512
numY = 512
numZ = 137
voxelWidth = 0.6 # set to 0.6 will trigger the bug
# voxelWidth = 0.7 # set to 0.7 is the normal case
voxelHeight = 1.26
offsetX = 0
offsetY = 0
offsetZ = 0
leapct.set_volume(numX, numY, numZ, voxelWidth, voxelHeight, offsetX, offsetY, offsetZ)
leapct.set_diameterFOV(640)
# Trouble-Shooting Functions
leapct.print_parameters()
# leapct.sketch_system()
g = leapct.allocateProjections()
f = leapct.allocateVolume()
leapct.set_FORBILD(f,True)
leapct.project(g,f)

When voxelWidth is set to 0.7, the projection looks fine. See the No. 4161 view below.

When voxelWidth is set to 0.6, the projection looks weird. See the No. 4161 view below.

[kylechampley]

Yep, you're absolutely correct, that is a bug. Thanks for the code to reproduce the issue. I found it and I am nearly done fixing it. I'll release a new version in the next 24 hours that resolves this.

[kylechampley]

This bug has been fixed and new code merged into the main branch. We also made a new release with this bug fix, v1.16.

When fixing this bug we also found that the scaling factor of the helical FBP algorithm was a little off and we fixed this.

Thanks for reporting this bug!

[oooo1114]

Thanks for your work!