The main graphical user interface for the EMAN package.
This is the main graphical user interface program for EMAN. It has 2 purposes. First, It will take you step by step through the process of performing a single particle reconstruction. Second, it has a file/history browser which allows you to analyze the results of your reconstruction, view image files, create figures for presentations, etc.
Important note: there are a large number of image display widgets in this program, there are also a few plot widgets. Any time you see an image or a plot, you can bring up an inspector (to adjust the brightness/contrast, print, save a snapshot, etc.) by pressing the middle button on the image or plot (or both buttons at once for those with 2 button mice). This is true in virtually every program in the EMAN package.
eman does not take any command line arguments. When you run the program, the above window will appear. There are 7 buttons in this window. The HELP and QUIT buttons should be fairly obvious.
The 4 steps
The top 4 buttons in the window represent the 4 stages of performing a single particle reconstruction in EMAN. When beginning a new reconstruction, 'cd' to the directory containing your raw data, and run eman from there. Next, press the 'step 1' button. This will cause a window to appear with a set of questions about the particle you are working with. Fill in all of the requested information. If you don't understand one of the lines, press the '?' button on that line, and a more detailed description will appear. Don't worry, none of this information will be sent anywhere. It will be stored in a file called '.form1' in the local directory for later use. When you've answered all of the questions, press the 'DONE' button.
EMAN will then examine the information you provided and generate a customized set of instructions for the first stage of the reconstruction. This will appear in a new window as soon as you press the 'done button'. If you wish to print out these instructions, they are stored in the local directory as 'form1.html'. This file can be opened later in a web browser, or you can get it to reappear by pressing the 'step 1' button again. All of your previous answers will already be filled in for you the second time. The current version of EMAN does not make as much use of the information you provided as it could, but in future versions, the instructions will become more and more detailed. Still, the answers are used, so be sure to answer all of questions accurately. When you finish step 1, move on to step 2, etc. If you are using the 'step 1' - 'step 4' buttons, make sure you run eman from the same directory each time. It will not allow you to run step 2 until you've completed step 1 in the current directory. The file/history browser
The other button, 'Browse Files/History' causes the file/history browser, pictured below, to appear.
This window is the main analysis tool in the EMAN package. It is fairly complex, and has a wide variety of features for image viewing and analysis. There are 3 sections to the main inspector window.
Section 1 is the file selector. It will contain a list of the files in the current directory. As a convenience, any files ending in '.hed' are hidden, so in IMAGIC file pairs, only one file will be listed. Directories are also listed, and single clicking is sufficient to change directory. In addition, you can change directory by entering a path in the text field above the file selector, or by pressing the 'Browse' button.
Section 2 is the history selector. This will contain a list of all EMAN commands (with a few exceptions) that have been run in the current directory. If a file is selected, the history selector will display ONLY history entries that are relevant to that file. To see a list of all commands again, select the directory tab in area 3. History Tab
Selecting an item in the history selector will cause the History tab to be selected in area 3. Section 4 will then display the relevant information about the line you selected, including the exact command that was run. The command string is in a selectable text box, so you can select it (left mouse drag) and paste it (middle mouse click) into an x-term window, editor, etc.
Changing directories, or selecting the directory tab will cause the directory inspector to appear in area 4. This inspector has some basic information about the current directory, as well as a comments window where you can (and should) enter a description of the files in the current directory. This information is saved in the README file in the current directory. This feature is provided to encourage users to label the contents of their image analysis directories. You never know when you'll come back to a project 9 months later and not remember exactly what you were doing in a particular directory.
Selecting an image file (single left mouse click) will cause the File inspector to appear (the file tab in area 3). This inspector displays relevant information about the selected image to the left, and a small view of the image itself to the right. If the image needs to be scaled down too much to fit into the small image display window, a message saying 'Big view required' will appear instead of the image. For multiple image files, the first image is displayed. In this case, you can scroll through all of the images in the file by changing the image number (starting with 0) to the left of the image. To view large images, click on the 'Detach' button. This will put the image view into its own window, which can be resized. To see all of the small images in a stack file at once, simply press the 'Matrix View' button. This will bring up a new window containing all of the images in the file, with scroll bars if necessary to see everything. Since all displayed images are loaded into memory, the program will ask you for confirmation if you try to open a matrix view containing more than 20 megs of data.
Image Control Panel (any eman GUI program)
(image of control panel goes here)
Pressing the middle mouse button over any image displayed in virtually ANY eman program will cause the image control panel to appear. This panel has quite a lot of functionality, and can appear with 2 variations. When displayed for a single image view, it will appear as above. If opened for a multi-image view, 2 additional check-boxes will appear. The 2 sliders at the top allow you to modify the display brightness and contrast. The 'Sca' text box allows you to modify the image scaling. There are 5 possible check-boxes:
- Invert - Will allow you to interactively invert the contrast of the image display. This can be useful for spotting particles in low contrast images.
- Label - This will place a numerical image number in the lower left corner of each image in a multi-image view.
- # Ptcl - This will display the 'nimg' parameter from the image header in the lower left corner of each image in a multi-image view. On a classes.*.img file, this number represents the number of raw particles used to construct a class-average.
- Hist - Display the image histogram. The bottom 1/2 of the view above will not appear unless this checkbox is selected.
- FFT - This only appears for single image views. It allows you to visualize the FFT of the displayed image. Note that the brightness and contrast are likely to require substantial adjustment.
Below the checkboxes are the mouse modes:
- app - This will pass mouse events on to the main application
- probe - This will allow you to use the mouse to probe around the image for specific pixel values. A magnified images of the pixels near the mouse cursor will appear in the lower right corner of the control panel. The numbers underneath the magnified view represent the position of the mouse with respect to the center of the image (as defined in EMAN). A red bar will also appear in the histogram representing the currently selected pixel, with the pixel value displayed numerically in the upper right of the histogram.
- move - This allows you to rearrange the images in a multiimage view. Simply drag one image over another, and they will be rearranged. Note that this has no effect on the file, but occurs only in the computer's memory. If you wish to save the rearranged image stack to a file, see 'save' below.
- split - This is a powerful feature which allows you to select individual images from a multi-image view. This can be very useful when preparing figures for talks/publications, etc. When you click on an image in 'split' mode, a new multi-image view will open, and a copy of the selected image will appear in the new window. You can continue to click on additional images. The new window also has it's own control-panel, etc. Note that the 'split' window cannot be closed. The only way to get rid of it is to close the parent image view that created it.
- delete - This allows you to delete individual images in a multi-image view. Note, again, that this change occurs only in memory. It does not directly affect the image file.
The histogram has several features to determine the value at a particular point in the histogram simply drag with the left mouse button and a red line will indicate the position under the cursor, with the value displayed in the upper right corner. To display image statistics, right-click on the histogram. The mean value is indicated as blue line. The green line indicates the 'center' of the histogram.
Three additional buttons may appear:
- Save - This will save the images in a multi-image view, complete with any modifications made by the browser, to an IMAGIC format image stack. Note that this routine will APPEND images to existing files. That is, if you save to the same file being displayed in the multi-image view, you will end up with the original images PLUS the images currently displayed.
- Save View - This will save a snapshot of the current image view to a PNG file. This is a lossless, portable file-format compatible with all modern web-browsers and office suites. This will save the view exactly as it appears on the screen, just like a screenshot.
- Print - This will print the current view to an available printer. Note that the printer support is broken in QT on several platforms. If 'Print' doesn't work properly, please use the 'Save View' button, then print the resulting image manually.
Compare Images Tab
When an image is selected, the 'Compare Images' tab will be enabled. When this tab is selected, a view similar to the one above will appear. This is the image comparison inspector. It is designed to allow you to compare 2 images within a single image file (a later version may allow comparison between images in different files). This inspector has a variety of useful features for image comparison. In 'General' mode, the individual image numbers to display on the left and right can be selected. The center image can then display a variety of different things:
- Difference - The images are subtracted and the (renormalized) difference is displayed.
- Aligned Difference - Image 2 is translationally and rotationally aligned to Image 1 before subtraction.
- Swap - Sometimes the best way to judge the differences between 2 images is to rapidly swap back and forth between image 1 and 2. In 'Swap' mode, pressing the 'GO' button will alternate between image 1 and 2.
- Animate - Currently does nothing at all.
CCF & RCF - These options are experimental and may even cause the program to crash. Avoid them for now.
You can also select the normalization to be applied to the image(s):
- None - The images are used exactly as they are in the file
- Edge - This will set the mean value around the edge of each image to 0 and set the standard deviation to 1.
- Std - This will set the overall mean image value to 0 and the standard deviation to 1.
- Pair - This will perform a least-squares fit of the pixel values in one image vs the values in the other image, then normalize one image to produce the optimal match. ie - the images will be normalized so the difference between them is as small as possible.
This inspector allows you to investigate the orientations represented by 2D images. It currently works in 3 different ways. In the 'Orient' mode, a single image (this may be a single particle, or a class average) is compared against a set of projections to determine how accurately its orientation can be determined. The comparison is made similarly to classesbymra, but no CTF correction is performed in this case. Select the file with the image you wish to orient, then select the correct image number at the left of the Euler inspector. The orientation determination will use whatever projections it finds in proj.hed/img, so make sure this file already exists with the proper angular spacing and symmetry. Pressing the 'Orient' button will begin the determination (it may take some time). When it is complete, a dot will be displayed for each projection in the triangle in the right side of the inspector. This triangle represents the asymmetric triangle of your particle with whatever symmetry you specified. That is, the top of the triangle represents looking at your particle along the Z axis, or north pole of your particle. The base of the triangle represents a fraction of a circle in the x-y plane (or above the plane for icosahedral symmetry). Each dot will be colored according to how well the selected particle matched that particular projection. The top 5% of values will be colored in gradually stronger red color. Below this point, brighter white indicates a stronger match. You can select a projection by clicking on one of the dots in the triangle. The selected projection will be displayed in the center window, with information below it (numerical quality factor, and orientation of the projection). From the width of the bright white area you can get an idea of the uncertainty in the orientation determination. Also, if there are multiple white areas, you can see if there are weaknesses in your model, like incorrect symmetry specification.
The second mode for the Euler inspector is 'Classes' mode. This is the most useful mode of the Euler Inspector. The triangle is used to display the distribution of particle orientations in a large data set. You must select a 'classes.#.img' file generated from a refinement run for this to work. When a classes file is selected, then the 'Classes' button is pressed, the triangle will display a dot for each class average in the file. The color will be determined by the number of particles that was used to generate each class average. In this case, the coloring is done on a logarithmic scale, so a doubling in brightness actually indicates 10x more particles in that class. Again, selecting an individual dot will display the corresponding class average and display the numbers associated with it. Note that if you did not use the euler2 option in refine, the class averages will be fairly evenely distributed in the unit triangle. With the euler2 option, the classes may be clustered in areas around the triangle. This is a good indication of preferred orientations in your sample.
The 'Classes' mode can also be used to display projections based on their position within the asymmetric triangle. If you select a 'proj.img' file in Classes mode, all of the projections will have the same color. Selecting one will allow you to see the projection appearance as well as it's Euler angles.
The third mode is 'Matrix' mode. This mode is used to view a similarity matrix generated by classesbymra in 'matrix' mode. It is still considered experimental. Convergence Plot
The final tool currently provided by eman is the convergence plot. To use this, you must run EMAN in a directory containing several completed iterations of a refine command, and optionally the results of a ttest command. This option will calculate the FSC (Fourier shell correlation) curves between sucessive 3D models generated by the refinement. Since each model is pseudo-independent from the previous model, this curve will give a good preliminary analysis of the resolution convergence of the refinement process. To get a more accurate resolution, a t-test should be run once the refinement has completed. This will generate 2 models using 1/2 of the data for each. This provides a fairly accurate estimate of the model resolution. Note that these 2 models are not completely independent, since the same reference model was used to classify each set of particles, but generally speaking, resolution will be noise/alignment limited, so this is not a serious problem. It may cause a slight overestimation of the resolution, should not be significant.
One trick worth noting: to display a legend of the (often confusing) array of colors and line types in this plot, hold down the control button and drag with the left mouse button in the plot to position the legend.
More details on using this plot are included in the 4 step reconstruction instructions.
Also note, that there is currently (1.2 (42)) a bug which causes eman to crash if you try to produce a second convergence plot within a single eman invocation.