Table 2 .

Main characteristics of the commercial 3D iterative reconstruction algorithms used in the study

Name	Type	Corrections			Noise regularization	Manufacturer default number of
		Attenuation	Scatter	Resolution		Subsets	Iterations
General Electric Evolution for Bone	MAPEM	From CT data, bilinear conversion of HU into attenuation coefficients at 140 keV	Jaszczak’s dual energy window method with 115 to 125 keV scatter window	Matrix rotation	One-step late method with green prior and median root prior at last iteration	10	2
				Row convolution with spatial resolution kernel stored in look-up table
Philips Astonish	OSEM	From CT data, HU segmentation using a step-like law, bilinear conversion of HU into attenuation coefficients at 100 keV, scaling to 140 keV	ESSE method	Convolution with spatial response function	Proprietary filtering (Hanning) of acquired projections and computed projections by forward-projection	15	2
Siemens Flash 3D	OSEM	From CT data, bilinear conversion of HU into attenuation coefficients at 140 keV	Modified triple energy window method with 108.5 to 129.5 keV scatter window	Matrix rotation	Gaussian post-filter (6-mm FWHM default value)	4	12
				Gaussian diffusion method with slabs

3D, three-dimensional; CT, computed tomography; ESSE, effective source scatter estimation; FWHM, full width at half maximum; HU, Hounsfield units; MAPEM, maximum a posteriori expectation maximization; OSEM, ordered subset expectation maximization.