AT SA - Automatic particle picking - statistical parameters ||

(12/12/97)

PURPOSE

This was written as part of the automatic particle picking program. This routine calculates statistical parameters for the data windows and determines the discriminant function to be used for future particle picking. Data windows should have bright particles over dark background. Various statistical parameters (viz., variance, skewness, entropy, kurtosis, some other secondary parameters introduced to enhance the local over global features (see note below.)are calculated for the input data windows (particles, noise and junk) and input to the discriminant analysis program as feature vectors which casts all the relevant information into a discriminant function.

SEE ALSO

AT WN [Automatic particle selection]
AT IT [Automatic particle picking - sort doc file]
AT PK [Automatic particle picking - peak search]

USAGE

.OPERATION: AT SA

.TEMPLATE FOR WINDOW SERIES: UNT****
[Enter the name of the data windows obtained from preparation part of the program and which have been categorized.]

.SIZE OF MINI WINDOW: 5,5
[Enter the size of the small window which is moved across the length and the breadth of the data window for evaluating local statistics.]

.DOCUMENT WITH CATEGORIES: DOC001
[Enter the document file containing two columns: the image number and the category that it belongs to (obtained after using AT IT). Remember that the categories that the program associates the numbers with are as follows:

 
           Category #		Category kind 
               1                    Particle 
               2                    Noise 
               3                    Junk 

.ANALYSIS RESULTS: ANALRES
[Enter the name of the file where you want to store the results obtained after the discriminant analysis is done and the discriminant function is created.]

.DISCRIM FUNCTION: DISCRIM
[Enter the name of the file where you want to store the discriminant function.]

NOTES

  1. The discriminant analysis program uses the discriminant function that it created to reclassify the input data windows and one can check how the function performs with respect to the initial assignment by you. In the "particle" category one includes data windows which contain single particles. In the "noise" category one includes data windows which contain noise. Very few of these windows are obtained if one uses a high threshold like 0.75 or so in the AT PK program. So in the first run or the TRAINING part of the program, one needs to use a very low threshold value in the peak search program in order to obtain more noise windows. In the "junk" category one could include stains, blobs, or clumped particles (particle aggregates).

  2. Details: autopartpick.html

SUBROUTINES

ATSA, TOMA, MEED, TIMA, POJ, FITT, STEP, DISC, COREL, PRP, DISTEST, MAHAL, MYWR0, MYWR1, MYWR2, MULT, MTNV, MTPR, DISCRIM, RYS1, RYS3, DIRNM, DIST1, DIST2, PRT, DIST4, HDIAG, MYREAD, FGR, HGR, INVBT, CHISQ, ALPHAINT, GAM

CALLER

UTIL4

© Copyright Notice /       Enquiries: spider@wadsworth.org