FindPoints

Overview

FindPoints searches a data set for peaks which exceed a threshold set in terms of an estimate of the data's standard deviation and mean. You can instruct the search process to automatically lower this threshold until a minimum number of points is found. To further constrain the search, you can load polygon data (typically from EditPolygon) and either have FindPoints assemble the polygons into volumes (objects) of interest or load an old-style object list from VolumeBuilder.

Once FindPoints has identified the peaks, you can also have it use the peaks as the starting point for region growing or for measurements of the points' intensity and their interrelationships.

Topics

Image Data

FindPoints requires either a Priism image file or window to determine the intensity at each volume element. If you use an image window, you can add or delete points found or select objects of interest using the right mouse button. See the section on picking for more details.

You can set which two dimensional cross-sections of the data set are used with the Z, T, and W controls on the main FindPoints dialog. For the z direction and time, the controls show the first section used, the maximum possible value for the last section used, and the amount by which the section index is adjusted when moving to an adjacent section whose data is used. Toggles allow you to set which wavelengths are used.

Topics

3D Objects

You can constrain FindPoints' operations by specifying volumes of interest (objects). To do so, you need to supply FindPoints with polygon data that represents cross-sections through the volumes; this data can either be in a file or, if you are working with an image window, the polygon data that has already been loaded into that window. EditPolygon can be used to generate either of these sources of data.

After loading the polygon data, it needs to be assembled into volumes. FindPoints can do this assembling for you if you press the Create button. When you do so, FindPoints works through the sections examining polygons in nearby sections; based on the amount of overlapping area, it associates those polygons into volumes. To limit what FindPoints considers a nearby section, set the max z gap; i.e. if you use a gap of 1, a polygon from section six in z can not be associated with a polygon in section eight unless there is a polygon in section seven which can be associated with the polygon in section six and the one in section eight.

The other option for assembling the polygons into volumes is to enter a filename into the object file field. The polygon file must already be loaded before entering this file name. The object file is an an existing file which specifies how the polygons are grouped. VolumeBuilder can be used to generate this file (it must be in the old-style format).

Topics

Process Mode

There are three modes for analyzing data with FindPoints. This setting will affect what portion of the data set will be used for the histogram statistics, peak finding and region growing.

Entire Data Set : is always used when there are no 3D objects. it can also be used when the data is broken down into 3D objects and all of the objects are to be processed at once. If region growing is done in this mode, the output mask data for all the objects will be written out to one image file, the size of the original data set.

Separate Objects: also processes all the objects in one run, but in region growing, a separate output mask file is written out for each object. In this case, the names of the files are the entered name, plus "-num", where num is the number of the object. This mode is helpful in speeding up the run time when a huge file is being processed, since breaking it down into smaller regions for processing will take up less memory.

Current Object: when there are 3D objects, a single object can be viewed in the histogram and the best set of parameters can be determined on a small region before processing the rest of the objects. When in this mode, either enter the object to analyze in the Current Wave, Time, Obj field or use the mouse to select the object in the window. See Picking.

Picking

The mouse can be used to pick in the image for one of three reasons:

Select Object: when there are 3D Objects loaded, the current object can be selected by clicking inside the object. The current object is only meaningful when the process mode is set to current object. When the object is selected, the polygons will turn blue and the Current Wave, Time, Obj field should update to the current object.

delete pts: in this mode, existing points can be deleted by clicking on them with the mouse

add pts: Points can be manually selected with the mouse. When this is done, the peak intensity in the local region is found. The local region is defined in using the search range fields in the peak finding menu. The box size is (2x + 1) of the search range. The point is then refined to the local center of mass if the Point Pos pulldown in the peak finding menu is set to Local COM or Local COM - Background. See Peak finding for more info on how the local center of mass is calculated.

Topics

Statistics

The Histogram menu displays a histogram of the data being analyzed, and shows three bars. The red bar is located at the mean of the intensities. the pink bar shows where the regions growing threshold is currently set and the white bar shows where the peak finding threshold is. The peak finding and region growing bars can be dragged to the position that seems appropriate according to the histogram. The value is reported as both threshold and multiples of the standard deviation in the histogram menu. Thresholds in FindPoints are set based on an estimate of the mean and standard deviation for the volume of interest. Three different methods can be used to estimate the mean and standard deviation:

curve fit: The mean and standard deviation are estimated by fitting a Gaussian to the histogram of all the values in the volume of interest.
overall std: The mean and standard deviation is calculated using all data points in the volume. These estimates are then iteratively refined five times by not including points whose intensities are two standard deviations greater than the mean.
dropoff std: The mean is taken as the center of the bin which is the peak in the smoothed (over five counts) histogram.

A fourth option is to just enter a threshold value directly. In this case, no statistics are done on the data. When this pulldown is set to Absolute thresholds, the peak finding and region growing menus' nstd input fields will be used for input of the threshold instead.

Topics

Peak Finding

The automatic peak finding algorithm uses pixels above the threshold value as starting positions for a local search for peaks. The threshold value is entered as number of standard deviations, or absolute threshold, depending on the Method setting in the main menu.

Once the list of peaks is determined, there are several criteria used to refine this list:

min num pts. If a minimum number of points is set and the toggle to auto decrease the threshold is set, the peak finding will be done over again, lowering the threshold by the number of stds entered until the minimum number of peaks is found. (note this function does not work if the absolute threshold is used instead of nstd).
min pt separation. If this value is set above 0, points will be eliminated if they are too close to a higher intensity point.
max num pts. If the resulting number of points is greater than the entered max, then the lower intensity points are thrown out.
point position. The final position for a peak can be one of four things:
Peak
The peak location is the center of the pixel with the peak intensity.
Local COM
The peak location is determined by the following steps:
1. Determine the origin for the center of mass calculation. When the Search Range values in the dialog are zero, the origin is just the center of the pixel with the peak intensity. Otherwise, a search is done by calculating the total intensity in boxes dimensioned 2 * xy search range + 1 by 2 * xy search range + 1 by 2 * z search range by 1 whose centers are no more than xy search range pixels in x or y and z search range pixels in z from the peak pixel. The center of the box with the largest total intensity is used as the origin for the center of mass calculation.
2. Calculate the center of mass for a box with dimensions xy box size by xy box size by z box size which is centered at the origin determined in the first step.
3. If the center of the nearest pixel to the calculated center of mass is not the origin, reset the origin to the center of the pixel closest to the center of mass. Repeat the second and third steps.
Local COM - Background
Uses the same procedure as Local COM, but when calculating the center of mass, the minimum intensity value for the pixels used in the center of mass calculation is subtracted from the other intensity values before calculating the center of mass. This is a way of compensating for a slowly varying background that would otherwise tend to bias the position estimate towards the center of the box used for the center of mass calculation. It may also help with errors in the position estimate caused by a box size that mildly truncates the tails of the peak. When these sources of systematic error are not significant, the Local COM method will give less noisy estimates of the position. If the box size substantially truncates the tails of the peak, both the Local COM and Local COM - background methods will have systematic errors and the Gaussian Fit option may give better results.
Obj COM (reg grow)
Until region growing is performed, the center of the pixel with the peak intensity is reported as the peak position. Once region growing is run, the peak position is reported as the calculated center of mass from the region grown about the peak pixel.
Gaussian fit
Fits the intensity values in the neighborhood of the peak to determine the center. The neighborhood is a box-shaped region with the dimensions shown in the COM/Fit Box fields. In cases where the fit is unsuccessful, the Local COM method (without iteration to reset the origin) is used to determine the center.

Topics

Region Growing

Region Growing starts from the points found with the peak finding or entered points, and grows out through the data using one of the following criteria to determine where to stop:

absolute intensity: A region is grown until the intensity drops below the value shown in the threshold value field. This option is only available when the method selected in the main Find Points dialog is "Absolute Thresholds".
std. dev. from mean: A region is grown until the intensity drops below the mean (background) level plus the threshold value field times the standard deviation. The mean and standard deviation can be seen in the histogram dialog. This option is only available when the method selected in the main FindPoints dialog is not "Absolute Thresholds".
half peak value: A region is grown until the intensity drops to halfway between the value at the peak (the seed for the region) and the mean (background) level. This is equivalent to using the "fraction of peak - mean" option with the threshold value set to 0.5.
fraction of peak - mean: A region is grown until the intensity drops to the threshold value * (peak intensity - mean) + mean.
fraction of peak: A region is grown until the intensity drops to the threshold value times the peak intensity.

To further limit how the region grows, you may specify the following:

xyz range: Regions are restricted to be within a box-shaped volume centered on the peak. The dimensions, in pixels, of the box in x, y, and z are twice the first, second, and third values, respectively, shown in the "xyz range" field.
stop at valleys: Growth of the region is stopped when the intensity values are observed to increase at further distances from the peak.

After the regions are grown, those which contain fewer pixels than the value shown in the "min pixels to accept point" field will be rejected.

The region growing process may generate two image data sets (if you use the "Separate Objects" processing mode in the Find Points dialog it is two data sets per object; in this case, the data set names have a dash and the number of the object appended). If a file name or window number is entered in the "Out Mask (opt)" field, a data set with that name will be generated. In it, the pixels that belong to a region will have an intensity equal to the region's number (a positive integer); all other pixels will have an intensity of zero. If a file name or window number is entered in the "Out Data (opt)" field, a data set with that name will be generated. In it, the pixels that belong to a region will have the intensity value from the input data set; all other pixels will have an intensity of zero.

Topics

Submit to queue

Once parameters are chosen for the peaks finding and region growing, the job can be submitted to queue. Currently, both peak finding and region growing are run when FindPoints is run in batch mode, and the result is written to a pick points format file.

Submit to queue - select the queue to run on; if "Write script only" is selected, the command and parameter files will be created but the calculations will not be started.

Command File - contains the environment setup and starts FindPoints with the parameter file, directing output print statements to a log file.

Parameter File - contains the parameters selected by the user for the current run.

Output Pick File - point information will go to this file.

Log file - contains parameters used and tracks progress of run

Topics

Results

Save Pick Points Format - output in this format contains the following information:
   Wave Time
    L BEAD # - 3DObject num. will be just 1 object if no polygons used.
    mean, standard deviation and threshold used for this object
    Point Info: x,y,z position, number, integrated intensity, total npts in region grow

Load Pick Points Format - an existing point file can be loaded back into FindPoints

Pt Data in Columns, Save Point Info - these are old, and no longer supported

Show Point Info - a table of point info on the screen. object num, point num, intensity.
table only show info for 20 points at a time so buttons at bottom let you look through
all the points. top starts at the first object, first point. next moves forward through the
list, and back move backward through the list.

Topics

Looping

Looping lets you quickly test a range of thresholds for peak finding and/or region growing to determine optimal parameters. The entire data set or one single object can be tested depending on the setting of the process mode pulldown in the main menu.

output pt list file - this is the name for the output pick points file. point positions and integrated intensities are reported in this file.

peak nstd start, inc - enter the starting multiple of the standard deviation to use as the threshold for peak finding and the increment multiple for the successive rounds. it makes the most sense to increment either the peak finding or the region growing but usually not both at the same time. set inc to zero to use the same peak finding threshold for each loop.

reg grow nstd start, inc - enter the starting multiple of the standard deviation to use as the threshold for region growing and the increment multiple for the successive rounds.

number of increments - the number of times to set a new threshold and run peak finding and region growing.

Get Looped - do it.

Topics

Defaults

Save Defaults - saves parameters on the main menu, peak finding and region growing menus to a file in
the .iveprefs directory. the file is called fp.par. filenames are not saved to the defaults file.

Load Defaults - loads the defaults parameters back in.

Topics

Overview

Topics

Related Priism Topics

Topics

Topics

Topics

Topics

Topics

Topics

Topics

Topics

Topics

Topics