Start by Selecting Models + TMV

Start the procedure by selecting one of the models you wish to use and tuning it up as described above. Make one or more manual mass measurements using that model. Repeat this for all models of interest. Be sure to include TMV (model 21) as one of the models. Basically you are "teaching" the program what to look for. Strike the <o> key to display the results to date. This enters all the displayed models into the search list. Select 'Delete *.smm' in the 'File' menu to clear all previous measurements. Next strike <F11> to activate the automatic search for all selected models. This masks the image at a threshold proportional to the height of the selected model above the background and traces around the outline of all isolated patches, determining maximum and minimum radius from the center. Particles approximating the size of the search model are marked with a circle or rectangle and entered in the *.smm list with category 'T' for trial. This procedure repeats for all models selected. If some particles were missed, you can enter them manually.

Test Auto-align on One Image

Clear the screen by striking the <0> (zero) key. Then hit <F12> to initiate model fitting. Again, particles passing the various tests will be indicated with a circle or rectangle and entered in the *.smm list, now with the category 'R' for realigned. This normally takes somewhat longer and some particles may be picked up by more than one model. Save the results by selecting 'Save *.smm' in the 'File' menu. Review by striking the <o> key again and move the highlight with the arrow keys.

Auto-align on Sequence of Images

If you are satisfied with the performance on a single file, you can extend the measurements to a block of contiguous files as follows. Move the mouse cursor to the small green bar of the last file in the sequence (bottom right panel). Its header should be displayed to the left of the green bars. Hold down the left mouse button and drag the mouse pointer to the magnification bar of the first image you want and release the left mouse button. All the bars of files of interest should be enclosed in the displayed rectangle and the starting and ending file numbers displayed. If this is not what you wanted, repeat the procedure. Images with the desired magnification values (inside the rectangle) will be analyzed when you hold down the <Shift> key and strike <F11>. The program will cycle through the entire procedure for each image, including the <F12> alignment step, with no further prompting. Depending on the number, size and complexity of the models and the speed of the computer, this may take 1-10 minutes per image and can be left to run unattended. The screen display gives some idea of progress. Note that the measurements for each image are saved in its respective *.smm file, so the automated analysis can include all specimens in a particular folder. Note also that the old *.smm results files are erased in this process, so if you wish to keep them you should rename them or move them to a different folder. The automated analysis run can be aborted by moving the mouse pointer to the bottom right corner of the screen if it is clear that there is a problem (note that there may be a delay while the current file is completed). If any portion of the measuring area goes off the edge of the image, the mass for that particle will be set to zero, but the other parameters will be saved. Measurements with zero mass are excluded from averages, although they are shown in histograms.

Check Auto-align Performance

The automated results for an individual file can be observed by reading in the desired image and striking <o>. Special care should be given to particles selected by more than one model, since they may be counted twice in some histogram protocols. If this is a problem, the models can sometimes be adjusted slightly to reduce overlap. If you move the mouse cursor to any particle on the left screen which has a previous measurement, the number of the model with the best fit and the fitting parameters will be displayed to the left of the zoomed particle image (overwriting part of the radial profile display). Parameters outside the selection range for that model will be displayed on a red background. It is also a good idea to check the parameter scattergrams on the right screen obtained by striking <o> and using the arrow keys to highlight the measurement of interest for several images. It sometimes happens that the first image in a series or the one used to tune up models is not typical. Most of the 'good' particles of each type should be close to the center in each scattergram panel and well within the selection ranges, especially the size and height panels. If this is not the case, you may be systematically removing large or small particles from the averages and skewing the results.

Average Specimen Parameters

Once you are satisfied that most of the particles of interest are being picked up and those failing are being rejected for good reason, you are ready to compute specimen averages. This is essentially the same as for the automated analysis, except that the rectangle should be drawn around magnification bars of only one specimen. Strike the <O> key (upper case O). The upper left panel will display file number, dose, background and abbreviated histogram for each model. The lower panel will show Ave and SD for each model as in the <o> mode, but for pooled data. The lower right histogram will show the distribution for measurements of the highlighted type. The data are presented in this way to focus attention on potential flaws in the data set.

Spot Skews and Trends in Averages

First notice any trends in dose or background value and see if they correlate with systematic variation in mass, particularly for the TMV. Mass loss is typically 2.5% at a dose of 10 el/A**2 normally used on a 0.512 micron scan. A 0.256 micron scan at the same beam current would deliver 40 el/A**2 and cause 10% mass loss, so it is usually best not to combine data recorded with different scan sizes without careful consideration. If the specimen contains detergent or salt, the background may fluctuate from image to image and the TMV M/L may fluctuate in the same or the opposite direction depending on the affinity of the contaminant for carbon film or protein. The TMV M/L should be 13.1 kDa/A with a SD of 2% or less under ideal conditions. If the particles are known to have a rigid shape similar to that of the selected model but many are failing the RMS test, the particles may be falling apart. That would also tend to show up as a skew in the overall histogram. For reference in judging this, a Gaussian with the same Mean, SD and integrated area is displayed in the main histogram window.

Results Files

Each time the <O> key is pressed two files are written in the c:\PCMass27\SpecAves\ folder. The first is called UXXXXXaves.dat and contains the Mean, SD and number of passing/number tested for each model and a simple histogram for the highlighted Model and type of measurement. The U number is the User specimen number of the first specimen in the analyzed sequence, read from the header (the number displayed below each black line at the bottom right). 'W' and 'H' numbers refer to Wall or Hainfeld specimens and controls. The second file is called UXXXXXparts.dat and contains a list of the particle parameters included in the averages. Note that pressing <O> a second time with the same starting file causes measurements to be appended to the previous file. This permits summarizing results for several models in the same file. We have attempted to include enough displays and analysis aids to give confidence in the reliability of the analysis and flag problems where the data or analysis may be flawed. However, we strongly recommend that new users discuss results with members of the STEM Group before publication. We are happy to provide optimized models and sample analysis whenever requested. A particularly effective approach is telephone contact with both parties viewing images of the same data and discussing the finer points of the analysis.

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