Dear David,
 
I've gained a little more experience deconvolving confocal data since the Proceedings of SPIE paper you cited.  For 2D deconvolution, the single most important element for a successful deconvolution is a satisfactory signal to noise value.  I have found that with the Nikon C1, a signal to noise of 12dB (usually requiring averaging somewhere between 4 and 12 frames per channel) always leads to a good result.  The deconvolution result you get from data averaged to greater than 12dB does not appear visually different from that from a 12dB image.  Noise leads to problems is mapping intensity data to its most likely source, probably because of errors in the psf estimation.  Remember that while noise is random, it is not smooth.  2D deconvolution works as well as it does in part because the spatial filtering of the confocal pinhole results in data that is very close to two dimensional.  While 2D deconvolution usually usually improves the appearance of single plane images acquired through a pinhole opened to 2 Airy disk diameters, a pinhole diameter closer ideal always leads to a better solution, especially when the data is averaged to 12dB.
 
3D deconvolution is also noise sensitive, but because data is brought in from planes above and below the nominal plane and the size of the data set used to produce the solution is larger, noisy data presents less of a problem.  Averaging to 8dB is usually adequate.   Given the choice between averaging three frames per step at a 0.3 um step size or acquiring the stack with 0.1 um steps without averaging, the higher axial sampling frequency usually leads to a better solution than simple averaging, probably because because while the size of the data is the same, the additional spatial information helps in arriving at an accurate deconvolution.
 
Regards,
Jeff Larson
Product Manager
Confocal Systems
Nikon Instruments Inc.
 

	-----Original Message----- 
	From: Confocal Microscopy List on behalf of David Biggs 
	Sent: Thu 12/2/2004 9:22 AM 
	To: [log in to unmask] 
	Cc: 
	Subject: Re: confocal pinhole size and deconvolution
	
	

	Search the CONFOCAL archive at
	http://listserv.acsu.buffalo.edu/cgi-bin/wa?S1=confocal
	
	Hi Jeremy,
	
	You may be interested in the paper by Jeff Larson referenced below that
	shows results for blind deconvolution of 2D confocal images with
	different pinhole sizes.
	
	Opening the pinhole will increase the number of photons and the haze,
	however, the signal-to-noise ratio will be improved, which benefits the
	deconvolution algorithm.  I don't believe it is necessary to collect a
	3D stack - just a single 2D image or 2D time-lapse sequence and run the
	deconvolution.  The extra photons that are collected will have come from
	close to the plane of interest.
	
	There is always a balance between signal and resolution, and if you have
	bleaching issues it would be better to sacrifice some resolution (which
	the deconvolution can attempt to recover) for improved signal quality
	(IMHO).
	
	Reference:
	
	        J.M. Larson, "2-D and 3-D deconvolution of confocal fluorescence
	images by maximum
	                likelihood estimation", Proc. SPIE 2641, 86-94 (2002).
	
	Please contact me off the list if you need any assistance processing
	your data.
	
	Cheers,
	David
	---
	David Biggs
	Senior Research Scientist
	AutoQuant Imaging, Inc.
	[log in to unmask]
	www.aqi.com
	
	
	-----Original Message-----
	From: Confocal Microscopy List [mailto:[log in to unmask]] On
	Behalf Of Jeremy Adler
	Sent: Thursday, December 02, 2004 7:33 AM
	To: [log in to unmask]
	Subject: confocal pinhole size and deconvolution
	
	Search the CONFOCAL archive at
	http://listserv.acsu.buffalo.edu/cgi-bin/wa?S1=confocal
	
	(1) Since confocal images are generally short of photons, the option of
	acquiring more photons by opening the pinhole and then deconvolving the
	image set to restore the quality, seems attractive.
	Given that image quality requires both resolution and contrast is this
	strategy viable ?.
	
	(2) Since deconvolution ideally requires a Z series that covers the
	likely size of a PSF and, if only single image is really required, is it
	better to acquire the Z series and deconvolve or to use the same
	acquisition time to improve the quality (acquiring more photons) from a
	single image plane ?.  This choice arises when the fluorophore of
	interest bleaches rapidly in single photon confocal imaging.
	
	--
	Dr Jeremy Adler
	Dev. Biol.
	Wenner-Gren Inst.
	Stockholm University
	S-106 91 Stockholm
	Sweden
	
	+46 (0)8 16 1568
	
	--