EmSART (cfg file)

The configuration file for reconstructing tomograms using EmSART is a textfile. Individual entries are separated by newlines.

The general format is

 Parameter = Value '''\n'''

Content

Parameter	Possible values	Description
CudaDeviceID	For a default four GPU setup: 0 1 2 3	List of the device IDs to use.
ProjectionFile	MRC image stack or DM4 file series with index in filename: file_01234.dm4	Filename with the projection images. Note: EmSART limits the number of interpolation steps why the original unprocessed tilt series is used here. I.e. not the aligned stack as produced by IMOD.
OutVolumeFile	File name with ending .em, .mrc or .rec.	Filename of the reconstructed volume.
RecDimensions	x-dimension must be a multiple of 4, y-dimension must be even, z-dimension is unrestricted.	Dimensions of the volume to reconstruct in voxels. X/Y correspond to the projection image plane at 0° tilt, Z is the thickness. The Y-axis is the tilt axis.
VolumeShift	X Y Z value	Volume shift with units “pixels of the micrograph”; this makes the shift values independent of the voxel size/binning. With VolumeShift [0] the origin of the reconstruction is in the middle of the Volume.
MarkerFile	EM file format	Filename of the markerfile containing tilt series alignment information. Format as used in “Clicker”.
Lambda	1	Relaxation factor used in SART. As we run only one iteration, Lambda has to be equal to 1.
Iterations	1	Number of SART iterations. Experiments showed, that best quality is achieved with Lambda = 1 and iterations = 1.
UseFixPsiAngle	false	determines if the markerfile contains in-plane-rotation angles for each projection image. If set to true, Psi-Angle must be set.
PsiAngle	Angle in degrees	In-plane-rotation image, if above is true, set the angle for all images
PhiAngle	0	beam declination (use with care, not well tested)
OverSampling	1-10 (more is possible but impact decreases)	Number of rays per pixel (for each dimension). See also the super-sampling SART paper. Also note that a voxel size > 1 corresponds to oversampling.
VoxelSize	1 = Same size as projection pixels; 2 = One voxel has the size of two pixels, etc.	Voxel size in pixels (the base unit in EmSART is "1 pixel")
CtfMode	True/false	enable CTF correction, if true, CTF file must be provided
CTFFile	Filename of the EM-file containing the defocus values	Defocus values for each projection.
CTFSliceThickness	Value in nm, usually in the range of the precision of defocus estimation, default is 50 nm	Thickness of a slice where defocus is assumed to be constant (= Precision of defocus gradient adaption)
CTFBetaFac	360 0 0.01 0	CTF correction parameters: first number gives the first maximum of the Thonrings in the power spectrum in pixels. No CTF correction is performed for frequencies below that value. The other three parameters (a, b and c) define the dampening envelope, where <math> env=e^{-ax-bx^2-cx^3} </math> with env the envelope function, a, b, c, the three coefficients and x the spatial frequency in 1/nm.
IgnoreZShiftForCTF	True/False	If False, it is assumed that the measured CTF from the CTF file is accurate in the middle of the volume (the z-shift is taken into account). If True, it is assumed that the measured CTF from the CTF file is accurate at the reconstruction origin (center of the volume + z-shift). In other words: If the reconstruction volume is shifted in Z (see VolumeShift), the defocus value can be adopted accordingly (tilt angle dependent) = Set to False If set to True, the Z shift is ignored and not compensated.
Cs	e.g. 2.7	Cs value of the projection lens, needed for CTF correction (in mm), can be omitted if CtfMode = false.
Voltage	e.g. 300	Acceleration voltage of the microscope in kV, needed for CTF correction, can be omitted if CtfMode = false
SkipFilter	True/false	Perform bandpass filtering on the projections or not before reconstruction
fourFilterLP	e.g. 1000	bandpass filter parameters in pixel: Low pass
fourFilterLPS	250	Low pass sigma
fourFilterHP	0	High pass
fourFilterHPS	0	High pass sigma
SIRTCount	1:15 (a number equal to the total number of projections would turn SART equal to SIRT)	Contrast = ordered subset size in super-sampling SART paper (1=perfect for sub-tomo-avg, 10-15 = good contrast)
BadPixelValue	X - Depends on micrograph pixel values	All pixels above X-times the mean value will be filtered out to cancel out X-rays
CorrectBadPixels	True/false	Correct for bad pixels or not
AddTiltAngle	Angle in degrees	Rotate the volume along the tilt/Y-axis by X degrees (not rad)
AddTiltXAngle	Angle in degrees	Rotate the volume along the X-axis by X degrees (not rad)
FP16Volume	True/false	Use 16Bit floats instead of 32bit to save memory
WriteVolumeAsFP16	True/false	If the one before is true, the result can be converted to 32bits before writing to disk
ProjectionScaleFactor	1 or 1000	Scale the images by 1000 (that's a good value...) if you use 16 bits. If 32 bit floats, use 1
ProjectionNormalization	“std” / “mean” / “none”	normalize the projection images by standard deviation or alternatively by the mean value to normalize the tilt series: <math> image = \frac{(image-mean}{std} </math> or <math> image = \frac{(image-mean}{mean} </math>. If “none”, no normalization is done, assuming an already processed tilt series, e.g. from IMOD.
WBP	True/false	false -> use SART not WBP. If set to "true", one can also use Weighted back-projection.
WBPFilter	Ramp/Contrast2/Contrast10/Contrast30	If WBP is set to true, one can change the WBP filter. The settings for Contrast2/10/30 corresponds roughly to the corresponding number of SIRT iterations for a dataset with 41 tilts with 3° increment. The parameters for the filter are experimental and fixed, this might change in future.
MagAnisotropy	1.016 42	Magnification anisotropy parameters. The virtual rays for forward- and back-projection are displaced elliptically where the larger radius has a stretch factor compared to the smaller radius as given in the first number at angle given by the second number. Note that the projection image is not interpolated for this compensation.
DownWeightTiltsForWBP	true/false	Normalization based on tilt angle. See https://www.sciencedirect.com/science/article/pii/S0969212615002798
PhaseFlipOnly	true/false	Use only phase flipping in ctf-correction, not Wiener filter
WienerFilterNoiseLevel	0.1	Used to adjust the threshold noise level for CTF correction. Should only be adjusted (increased to ~0.2) for high resolution reconstructions/alignments (< 5A).

Example

Note: All values that are typically changed case-by-case are left empty.

    CudaDeviceID = 0 1 2 3
    Lambda = 1
    Iterations = 1
    ProjectionFile = 
    OutVolumeFile =  
    MarkerFile = 
    CtfFile =  
    RecDimesions =  
    VolumeShift = 
    VoxelSize = 
    AddTiltAngle = 
    AddTiltXAngle = 
    UseFixPsiAngle = false
    PsiAngle = 
    PhiAngle = 
    OverSampling = 
    CtfMode = true
    CTFBetaFac = 
    Cs = 2.7
    Voltage = 300
    IgnoreZShiftForCTF = false
    CTFSliceThickness = 50
    SkipFilter = true
    fourFilterLP = 0
    fourFilterLPS = 0
    fourFilterHP = 0
    fourFilterHPS = 0
    SIRTCount = 
    CorrectBadPixels = true
    BadPixelValue = 10
    Crop = 50 50 50 50
    CropDim = 10 10 10 10
    DimLength = 50 50
    CutLength = 10 10
    FP16Volume = false
    WriteVolumeAsFP16 = false
    ProjectionScaleFactor = 1
    ProjectionNormalization = std
    WBP = false
    MagAnisotropy =