There is optical analysis software that can do complex raytracing and calculate this, but we’ve always had an optical engineer who spent most of their days designing lenses for that level of complexity.
It’s hard to tell the desired effect and complexity from the photo you posted, but in general good light dispersion requires - the right LED (with a cone that matches the intended illumination area), the right plastic to either diffuse the light (we used a lot of Lexan FXD materials for light pipes since they could be milky to diffuse light better), and the right distance from the PCB for the cone to spread to hit the intended area.
The easiest way to get a rough estimate of this working is to find the cone of your LED (ex 60 degrees) and model it into your CAD model so it’s shining out of the LED. If the LED is shining straight towards the surface that will be fairly easy. You want the positioning of the LEDs to be enough where you have a slight overlap of each cone (since there is falloff from the center to the outside of the cone).
If you can get the LEDs laid out on the board where they all consistently intersect with the surface you want to light up, then you likely don’t need any “pipe” features at all. If the LEDs are at an angle or awkward position, then you would have to look at actually building a pipe to try and route the light up which is a bit more complex.
I’ve done experiments using Keyshot raytracings to try and simulate some of the effects, but generally the renderings take forever and are still only so accurate. Once you have a rough idea, then you can move to a 3D printed part (it may need to be hand polished) or a cast urethane part.
If you want a better idea of the approach to use and can share a cross section of where your PCB is and where you want to illuminate its sometimes easier to sketch.