some potential errors in algorithm implementation

[Yuliang-Huang]

several points:

1. In the original collapsed cone paper (Ahnesjo, 1989, Medical Physics 16(4)), "to avoid jittering, (cone) lines must be defined such that all voxels are passed once, and only once by each direction" of the cones. But I did not see checking of this condition in your kernel convolution code.
2. Fluence should be calculated by transporting the surface fluence to the voxel through effective depth and inverse square law. But your code simply used inverse square law.