Local Background

Local background is measured as follows. In the circle mode, three concentric circles are drawn. The pixels within the center circle are summed for the mass measurement. Pixels outside circle 2 (5 pixels larger than circle 1) and inside circle 3 (twice the radius of circle 1) are used to compute the local background. Note that the background mask is retained and only unmasked points are used for local background. As you move the mouse, unmasked pixels between the outermost two circles appear with normal contrast while masked points change to white. Excluding these points may result in no useable pixels, in which case the picture background is used for L_Mass also. In rectangle mass, larger concentric rectangles are drawn, similar to the circles discussed above. For rectangle M/L (mass per unit length) mode bands are drawn on both sides of the measuring rectangle and the local background computed in the outer bands on either side of the centerline. Using the background mask makes it possible to get a useable local background even in crowded areas.

Picture Background

The mass using the picture background, P_Mass or Mp, is generally more reliable and gives a better standard deviation (SD) than the mass computed using the local background. L_Mass or Ml may be interesting in special cases and any significant difference between L_Mass and P_Mass is a sign of trouble in the measurements. Model mass will be discussed below.

Mass Measurement

Make a mass measurement by positioning the circle or rectangle over a particle. The contents of the measuring circle or rectangle are zoomed or de-zoomed to fill a viewing window of 80x80 pixels below and near the right edge of the Ch0 image. The P_Mass or P_M/L is displayed immediately under the zoom window and the L_Mass or L_M/L on the next line down. At this point it is instructive to vary the radius of integration (<t>/<y>) and observe how the mass value changes. If the radius is too small, the measured mass will drop. As the radius is increased from a too low value, the mass should increase, then plateau, and finally begin to diverge as extra background noise is included. Move to "clean" background areas and note the fluctuation in mass from one area to another (should be symmetrical above and below zero).

Recording Mass Measurements

Record a mass measurement by centering the measuring circle or box and pressing the left mouse button or striking the <=> or <ENTER> key (slower but less likely to move the center point). "Fine tune" the centering by leaving the mouse fixed and moving in single pixel increments with the arrow keys. A circle, rectangle or pair of lines should be drawn around the particle when it is entered and the sequential measurement number should appear next to the particle, followed by the value of P_Mass. The measurement parameters and mass values are saved in a buffer, which is written to disk when a new image is read or the program is shut down. The LA particle image is also copied to the right screen, along with the reference model (see below), difference image with superimposed rotational power spectrum, SA particle image and LA minus SA image (may be non-zero on salt contaminants). Particles can be deleted starting with the one most recently entered by striking the <Backspace> key. If a previous measurement was made close to the particle, that will be displayed for your information. List the measurements to date by striking the <i> key (repeat <I> for more measurements, clear the screen by moving the mouse).

Reference Shape Models

A reference model is displayed in the small panel below and to the right of the zoomed measuring area. This serves several convenient purposes. Mass measurements are labeled with a "reference model" number and 80 separate running averages are maintained. This permits measuring different types of particles in any order without scrambling the averages. One of the 80 available models may be a reasonable approximation to the size and shape of your object, so it provides a convenient visual reminder of which type of particle you are aiming for. Each has a pre-programmed radius of integration, length, size, shape and name. Selecting that model loads the respective parameters. To select a model, choose "select" from the "Model" menu, hit the <m> key or click the mouse on the model display panel (bottom center). Move the mouse until the model of interest is highlighted and press the left mouse button. Repeat this until you find the one you want and adjust radius of integration etc. to fit your particles.

Difference Image

The area to the right of the zoomed image and above the zoomed model displays the difference image. A search routine adjusts the model parameters to minimize the RMS (root mean square) difference between model and image pixels. The numerical results of the optimization are printed to the right of the difference image and the parameters describing the quality of fit are printed below the mass measurements. Any parameters outside preset limits are printed on a red background. DX and DY are displacement in pixels from center pointed to by the mouse. T is the rotation angle of the model. SZ is model zoom giving the best fit and HT is the relative signal amplitude. The quality-of-fit parameters are normalized so that a perfect fit would give a value of 1.0. Those three parameters are: (1) brms, the RMS deviation from background of pixels between the inner circle and the middle circle (or rectangle), (2) rms, the RMS of image minus best-fit model for all pixels within the innermost circle and (3) srms, a comparison of rms values with the model at the best rotation angle versus that with the model at the worst rotation angle. The upper and lower cutoffs for each selection parameter are given in the model definitions. For advanced users these reference particles can be used for automatic particle selection and alignment, as in PCMass15. Models are defined in the PCMmods.dat file supplied and are easy to customize as described later. The color of the particle number:Pmass value printed next to the particle is keyed to the current reference model number. Note that the measuring and model windows both zoom as the radius/length of integration are changed. Note also that a rotational power spectrum is displayed at the bottom of the difference image, indicating the most prominent symmetry of the particle in the viewing window [Not available until bkg is computed].

Center of Model Particle

The "best" center of the particle as found by the refinement program is printed below the srms parameter. The coordinates are in pixels relative to the lower left corner of the image. This is the first quadrant of a Cartesian coordinate system and angle increases counter clockwise. Note that the computer screen coordinate system has 0,0 in the upper left corner and this may lead to some confusion; it certainly is an annoyance in programming. The starting center is the sum of the mouse X,Y printed below header plus the keyboard offset from the arrow keys, printed above the mouse X.Y plus the DX,DY values determined by fitting. The image intensities on channels 1 and 2 are printed after the mouse X,Y. You can check the intensities of neighboring pixels using the arrow keys. The XKB and YKB values are reset to zero by a mouse move greater than 2 pixels.

Test Example

After reading in image c:\nk7375\nk737537, running the background program, selecting model 50 (default), and moving the mouse to X=299, Y=60, the bottom center of the screen should present the following information: |************||***********| DX=0.270 model X offset in pixels | Zoomed | Zoomed | DY=0.318 model Y offset | Image | Difference | T= 14.104 model rotation angle | | Image | SZ=1.008 model zoom | | | HT=1.013 model relative intensity ************|*********** | VS=29% variable subunit in center Mp=3.660 | Zoomed | Ml=3.695 | Model | R=16 L=0 radius of integration, rectangle length brms=0.743 | | (in pixels) rms=0.641 | | srms=0.327 *********** 299.3, 60.3 50 Ehb face Model X,Y model number and name Mm=3.466 model mass after fitting (Note, entries on a red background are outside selection parameters for that model.) Mp = particle mass or M/L using global (picture) background Ml = particle mass or M/L using local background brms = RMS deviation from background for pixels between circle 1 and circle 2 rms = RMS of image minus model for pixels inside circle 1 srms = a measure of the best fit versus that with the model rotated to the worst position X,Y is the fitted center in pixels relative to the lower left corner of the screen

Processing Sequence

Depending on the speed of your computer, you may notice a delay in completing this information. To be compatible with Windows multitasking, PCMass processes the image in short bursts. The first priority is moving the mouse pointer (arrow) in response to mouse movements. The next operation is filling the image zoom window, followed by drawing circles or rectangles, fitting the model, redrawing if necessary, then filling the zoomed model and difference image windows. Finally, the program checks for previous measurements near that point with the selected model, printing the relevant parameters to the left of the zoomed image area. If the mouse is moved at any time the sequence is aborted and starts over at the new image point. The arrow keys allow fine-tuning the mouse pointer in any direction.

Results: Model Fitting

Any delay caused by model fitting is compensated for by elimination of the need for rotational alignment of rectangles (eg TMV). The model zoom parameter, SZ, gives a quick measure of relative size. If the model is a reasonable approximation of your object, the difference image highlights imperfections or accretions. The fitting process can be viewed at several different levels using the DISPFLAG parameter at the bottom left of the screen. Normally this is set to zero for the fastest operation. Increasing the level increases the detail of the displays provided. Many of the displays in levels 3 and 4 pause after certain operations (to allow careful study) until you move the mouse.

Results: Background

The value of picture background is obtained by bilinear interpolation from the 16 values calculated by the background program. The dose is calculated from the background standard deviation and absolute intensity, assuming this is mainly due to counting statistics. brms is computed in the 5 pixel wide swatch just outside the radius of integration and is the ratio of bkgSD from the background computation divided by root mean square (intensity minus bkg). For clean carbon, this should be 1.0. If the radius of integration is too small or there is dirt around the particle, the value of brms will decrease rapidly. A value less than 0.5 is a sign of serious problems with that measurement. If the background has not been run, zero will be printed for this parameter.

Results: rms and srms

The absolute value of the rms parameter is normalized relative to the dose and average signal within the measuring area and should be close to 1.0 for particles which can be fit well by the model. If the background has not been run, zero will be printed for this parameter. The absolute value of the srms parameter is rms(aligned)/{rms(aligned)+rms(rotated)}. The rotation angle is half the rotational symmetry angle of the current model. For objects with large rotational modulation such as TMV, values of srms of 0.9 or greater are possible (90 deg rotation). Typical earthworm hemoglobin values are 0.4 (30 deg rotation). An interesting exercise is to move the measuring circle or rectangle over clean background and particles, noting the variation in mass values. Then center on a particle and vary the radius of integration (<y>/<t>). If the radius is too small, part of the mass will be missed and if it is too large, the value will fluctuate over an increasingly wide range as a larger background mass (as well as stray objects) is integrated, then subtracted out imperfectly. Proper choice of the radius of integration is essential to good mass values.

Results: Projected Mass Profile Curves

The panel to the left of the zoomed image displays the projected mass profile. For filaments, this is the average signal as a function of distance from the centerline and for round particles it is the average signal as a function of radius from the center of the zoom window. The blue curve uses large angle data and green is small angle. The red curve is the apparent density of a cylindrically or spherically symmetric object which would give the observed projection. Note that left and right halves of the curve are computed independently so asymmetry and noise are more evident. Note also that for round particles the center of the curve is noisy due to the smaller number of pixels contributing to the average.

Updated 4 Oct 2007	Security Notice   Webteam   Site Map   STEM   Biology   BNL