Someone'll need to correct me on these if I am wrong.
The bits are to do with colours. 24bit is 'true-colour', 32bit is evil and rarely, if ever supported (made worse by some programs choosing to store it as XBGR and others as BGRX), and 8 and 16bit (the latter of which is also rarely used but still more common than 32bit) use palettes.
A PNG is usually 32-bit, but since that should remain out of the question for various reasons, to maintain the best colours 24bit is the best thing to use, though you lose the alpha channel meaning transparency must be done via chroma keys and so-on.
You could always track down a .dll which provides .png-loading functionality...or learn a real language.