ROI stands for Region of Interest. An ROI designated a volume in space over which statistics should be calculated.
The following ROI types are currently supported in AMIDE:
An ellipsoid is similar to a sphere, but with a diameter specified for each direction [x,y,z]. In the case of x=y=z, the ellipsoid is a sphere.
An elliptic cylinder is similar to a regular cylinder, except it has an ellipse as its base instead of a circle.
Exactly what it says, a 3D box.
Isocontour ROI's are regions selected from the data set such that the edge values of the ROI are always the same value. There are two types, 2D and 3D isocontours. Additionally, both these ROI types can be defined so that they encompass all neighboring values either above a certain minimum value, below a certain maximum value, or between a minimum and maximum value. After drawing of these ROI's, they can be modified by using manual drawing or erasing operations.
A 2D isocontour is derived by considering a value on one of the displayed 2D slices. The depth of a 2D isocontour is specified initially by the depth of the viewed slices.
A 3D isocontour is derived by considering a value on the current frame of the active data set.
Freehand ROI's are regions of interest that are drawn manually.
A 2D Freehand ROI is similar to a 3D Freehand, except that it is constrained to be only one voxel thick. By default, the depth is the current slice thickness, although this can be changed by the user.
An ROI that can be drawn freely in all 3 dimensions..
To draw an ROI, you first need to create a new ROI. You can add a new ROI to either the study, or a particular data set. To add an ROI to the study, you can either select the ROI desired under the "Edit->add ROI:" menu item, or right click on the blank area of the study tree. To add an ROI to a data set, shift-right click on the data set that you'd like to add the ROI to. In both cases, a dialog box will pop-up for you to enter in the new ROI's name.
When first added, the new (undrawn) ROI will be selected in the study tree. When an undrawn ROI is selected in the study tree, the program will use the next mouse input on any of the displayed views to begin the process of drawing this ROI.
For ellipsoid, elliptic cylinder, and box ROI's, a click with the left button will begin an edge-to-edge drawing, while a click with the middle button will begin a center-out drawing. The x and y dimensions of the ROI are determined by this process. The z dimension (thickness) of the ROI can be specified by the pop-up dialog that will appear on the completion of the mouse movement.
For isocontour's, the value of the data set at the clicked upon location will be used to derive the isocontour.
For freehand's, the point on the screen that is clicked upon will be included in the ROI.
After an ROI is drawn, it can be further manipulated to adjust its size, placement, and orientation. You can directly manipulate the ROI by clicking on it in any of the viewing windows. Mouse button 1 is used to shift ROIs. Mouse button 2 is used for zooming ellipsoid, elliptic cylinder and box ROI, and is used for entering drawing mode for isocontour and freehand ROI's. Mouse button 3 is used to rotate ellipsoid, elliptic cylinder, and box ROIs, and for redefining the isocontour value for isocontour ROI's.
For isocontour and freehand ROI's, drawing mode can be entered by using thie middle mouse button (button 2). Once entered, points can be added or removed from the ROI by using the left (button 1) or right (button 3) mouse buttons, respectively. Holding down the shift key while using these buttons increases the size of the action. The middle button (button 2) allows the user to leave drawing mode.
You can also edit the ROI size/placement/orientation/name etc. by clicking on mouse button 3 while over the ROI's name in the study item list. This brings up the ROI modification dialog (described at the section called “ROI Modification Dialog”).
Statistics on an ROI can be calculated via the "Tools->calulate ROI statistics" menu item. Choosing this will pop-up a dialog that lets you choose which ROI's (selected or all) and which data sets (selected or all) you'd like to calculate statistics over. You will also have three options as to how you what the values to be calculated.
Calculate over all voxels.
Calculate over highest x percent of voxels. For example, if you choose this and pick 25% as the number, your ROI will be calculated from the 25% of the voxels in the ROI that have the highest values.
Calculate for voxels >= % of Max. This method is based on Lee, Madsen, Bushnel, and Menda, Nuc Med Comm 2000, 21:685-690. As an example, if you choose this and pick 50% as the number, the highest valued voxel in the ROI will be found, and then the ROI statistics will be calculated for all voxels that are greater or equal to 50% of the highest valued voxel.
Calculate for voxel >= Value. This algorithm only does calculations for voxels in the ROI that have a value greater than the value specified.
There's also a check box to enable "more accurate quantitation". The default algorithm (corresponding to unchecked) makes some approximations in deciding which voxel are in our out of the ROI. If this check boxed is checked, the ROI results will be more accurate, but will take much longer to compute.
After hitting execute, the program will crank for a while, and then show the calculated values in a new dialog window. Hitting "Save as" button allows saving these values as a tab separated values (TSV) file. This file should be easily imported into most spreadsheet applications (Excel's a little stupid, you may have to explicitly tell it you're importing a TSV file). Pressing the "Copy" button copies the information into the operating systems clipboard, allowing pasting of the results into other programs. The "Save Raw Values" button allows you to export the underlying raw data values for the ROI's in case you wish to do your own statistical analysis.
Currently, AMIDE generates variance and standard deviation values that may occasionally be of interest to imaging physicists. It is very important to remember, that these numbers represent the noise in the data set, NOT the noise in your experiment.
The variance of an experiment can only truly be measured by taking multiple samples (i.e. performing multiple scans) and calculating the variance between these different samples.
Short story: The calculated volume shown by the ROI statistics dialog is correct. Use this value as the volume of the ROI, not the value you might calculate by hand based on the ROI's dimensions.
Long story: AMIDE calculates ROI's by translating the ROI's dimensions into the data set's coordinate space. It then computes statistics for all the data set voxels that are in the ROI. For voxels that lie on the edge of the ROI, AMIDE will subdivide the voxel into a finite number of subvoxels, and calculate over the subvoxels. This approach yields correct statistics, but it is important to realize that the computed ROI is a discrete representation of the specified analytical ROI. So while the true volume of an ellipse is pi*r1*r2*r3, the computed volume of the ellipse in AMIDE will depend on the number of voxels and subvoxels that were determined to lie within the ellipse, which in turn can depend on the orientation of the ROI with respect to the data set in question. Since the computed volume given by AMIDE represents the volume in the data set that was used for the ROI calculation, you will want to use that value (not the real ellipse value).
AMIDE doesn't present the "total" value in the ROI, as it doesn't necessarily know what the units of the underlying data are. If you're using PET or SPECT data, your voxel values are most likely proportional to activity/volume/time. To calculate the total in your ROI, you should multiple the mean value of the ROI times the ROI volume and the frame duration. If you're using CT data, your values are probably proportional to density, so to calculate the total you would multiple the mean ROI value by the ROI volume.
This is the median value of all the voxels that are enclosed (partially or totally) within the ROI. For an even number of voxels, the median is defined as the average of the center 2 values.
The mean value of the voxels in the ROI. Voxels that are partially enclosed within the ROI are appropriately weighted.
The variance of the voxels in the ROI. This is a weighted variance calculation so that voxels that are partially enclosed within the ROI are correctly handled.
The square root of the variance.
The square root of the variance, divided by the square root of the total number of voxels in (totally or partially) the ROI.
The minimum and maximum values for all voxels enclosed totally or partially within the ROI.
The volume of an ROI (mm^3). Details as to its calculation are above in: the section called “Changing Calculated Volume”.
The is the sum of the voxel weights, and gives an indication of how large the ROI is in voxel space.
This is the total number of voxels used in calculating the ROI, both partial and total. In contrast to the "Fraction Voxels" measure, the "Voxels" measure gives a better indication of the statistical validity of the mean, variance, etc.
To directly modify parameters of an ROI, right click on the name of the ROI in the study tree to pop-up the modification dialog. Parameters that can be modified are divided into the following pages.
The name and type of ROI can be altered on this page.
The center of the ROI can be shifted with respect to the origin on this page. The x, y, and z parameters are in millimeters.
The size of the ROI can be altered from this page. The x', y', and z' dimensions are in millimeters and are orientated with respect to the orientation of the ROI.
The ROI can be rotated around its center in this page. There is one dial for each of the three slice planes. The transverse dial will spin the ROI in the transverse plane (i.e. rotate on the z-axis). The coronal dial will spin the ROI in the coronal plane (i.e. rotate on the y-axis). And the sagittal dial will spin the ROI in the sagittal plane (i.e. rotate on the x-axis). The "reset to default" button allows the ROI to be rotated back to the default orientation. On the bottom of this page is a matrix showing the coordinate frame of the ROI with respect to the base coordinate frame.