Scanner coincidence window depends on TOF data

[KrisThielemans]

We currently cannot set the coincidence window, even if we know it, or it is contained in the data (for instance, for Siemens data we ignore the field, and for GE data we don't read it either). The current code is

STIR/src/buildblock/Scanner.cxx

Lines 2214 to 2218 in 8c57bc9

	Scanner::get_coincidence_window_width_in_ps() const
	{
	const auto w = this->get_size_of_timing_pos();
	if (this->is_tof_ready())
	return this->get_max_num_timing_poss() * w;

with

STIR/src/include/stir/Scanner.inl

Lines 263 to 266 in 8c57bc9

	Scanner::is_tof_ready() const
	{
	return (max_num_of_timing_poss > 0 && size_timing_pos > 0.0f && timing_resolution > 0.0f);
	}

This therefore only works for TOF data. (It is also potentially dangerous if someone would create a wrong scanner object with fewer (uncompressed) TOF bins than actually used, but that seems up to them and could be addressed via documentation)

The coincidence window is currently only used in the RFS calculation

STIR/src/data_buildblock/randoms_from_singles.cxx

Line 37 in 8c57bc9

coincidence_time_window = scanner.get_coincidence_window_width_in_ps() / 1e12F;

. This is currently only used for GE data (but it would therefore fail when using trimmed data, or for the Omni as that's non-TOF).

After thinking about this a bit, my impression is that we don't need a separate coincidence_window member in Scanner, as even for non-TOF, we can say the num_timing_poss=1 and set the TOF bin width to the coincidence window. We can therefore already set it in the Scanner initialisation as well as parse this from the header etc.

Of course, it's going to be hard to find timing resolution for non-TOF scanners, so the condition on is_tof_ready would have to be changed. This is already done in #1430:

STIR/src/include/stir/Scanner.inl

Lines 263 to 266 in 34a35de

	Scanner::is_tof_ready() const
	{
	return (max_num_of_timing_poss > 0 && size_timing_pos > 0.0f ); //&& timing_resolution > 0.0f);
	}

In any case, it'd be nice to check the computed value with whatever the scanner reports (if it does).

@danieldeidda if you know the relevant info for the Mediso AnyScan, you can already add it in your Interfile headers, and (ideally) Scanner.cxx.

@NikEfth are we missing something else?

[danieldeidda]

After thinking about this a bit, my impression is that we don't need a separate coincidence_window member in Scanner, as even for non-TOF, we can say the num_timing_poss=1 and set the TOF bin width to the coincidence window. We can therefore already set it in the Scanner initialisation as well as parse this from the header etc.

I think I am confused, then what do I set it to if there is no member? also what is the keyword to save into the header?

[KrisThielemans]

If you'd fill in

 Maximum number of (unmashed) TOF time bins := 264
 Size of unmashed TOF time bins (ps)        := 17.9039
 TOF timing resolution (ps)                 := 214

(these are values for the Vision 600), it'll compute the coincidence window from the above as per the code.

[danieldeidda]

So are you saying to have these even with scanners with no ToF?

[KrisThielemans]

yes, maximum number, size of bin = coincidence window, tof timing resolution: something very big. (not used for the coincidence window calculation)