1. Field of Invention
This invention relates to a method and system for automatically processing images from a micro-array using an image processor. More particularly, this invention relates to a DNA micro-array image segmentation algorithm with statistical analysis of spot centre overlay pattern, that automatically segments sub-grids and spot positions of genetic micro-array images and is used to further obtain the intensity of the spots using the overlay pattern of spots.
2. Description of the Prior Art
Micro-array image technology is a powerful tool for studying the expression of thousands of genes simultaneously. The micro-array images are obtained with different image quality making automated analysis of the micro-array spots very difficult. Some micro-array images are of a high image quality, i.e., the spots in the images are very clear without any noise in the image background. However, micro-array images with inferior image quality have some noise. Thus, the task of automated segmentation of the spots accurately and consistently must be immune to background noise as well as poor image contrast of the spots with respect to the background.
In Shams U.S. Pat. Nos. 6,349,144 and 6,577,956, there is described an automated DNA array image segmentation analysis method that is not completely automatic and has several limitations. In column 6, beginning at line 1, when a computer is used, the user selects an image file for processing, stores the image frame and displays on the display as a control image. The user is then said to select an image region and further to specify a number of columns and rows of arrayed image spots. Further, the imaging and software system described in the Shams Patent assumes that the pixel intensity corresponding to the DNA spots are greater than their surrounding background intensity values. Further, the method requires storing a frame of image information in a memory device and generating a grid in the memory device. FIG. 8 is said to illustrate an example of a general flow graph diagram for the program instructions of the computer system and the software system. The program instructions include at least two manual steps by a user in obtaining the four corners of an image area and indicating a satisfaction with grid placement.
With the large micro-arrays that are available and the large time input required for each manual step in any analysis or other processing of images, it is extremely important to provide a fully-automated method and system. Unfortunately, the systems and methods described in the Shams Patents are only partially automated and require steps that require a large time input and steps that can be eliminated entirely.
The goal of the algorithm is help automate the analysis of images produced by gene array chips. In developing this type of tool, the first task is to provide a software technique to automatically identify regions of the micro-array image, identify the spots in the image and obtain image intensity information to be analyzed and displayed by another algorithm.
In achieving the first task of automated image segmentation, the algorithm uses mathematical morphological operations and image measurement techniques. In achieving automated image segmentation, the algorithm proceeds along the steps listed below:
A method of automatically processing images from a micro-array using an image processor, the method comprising pre-processing images to ensure that the images meet a predetermined threshold, inputting images that meet the threshold to the image processor, the image processor automatically calculating a size of a spot image, automatically calculating a spacing between adjacent spot images, automatically generating a first grid, automatically adjusting the first grid to fit the spot images being processed and reporting an output.
A system for automatically processing images from a micro-array comprises a microprocessor controlled by software, said software controlling pre-processing of the images to ensure that the images meet a predetermined threshold. The images that meet the threshold are inputted to the microprocessor by the software. The microprocessor is then controlled by the software to automatically calculate a size of the spot image, automatically calculate a spacing between adjacent spot images, automatically generating a first grid, automatically adjust the first grid to fit the spot images being processed and to report an output.
These and other features, aspects and advantages of the present invention will become better understood with regard to the following description, appended claims and accompanying drawings where:
1. Overview
This invention provides a fully-automated segmentation technique and its application software for automated segmentation of the spots in micro-array images. This invention can be used to automatically identify the grid overlays of micro-array images. The grid information can be used to initialize the positions of the spots, and then automatically calculate the intensity of every spot position within a certain area. This invention can be used effectively for low-noise and high noise micro-array images. This invention includes optimal image processing analysis, spot size analysis, noise reduction, and grid statistical analysis.
2. Operations of Embodiments
At step 106, the intensity histogram of the image is checked, and then the minimum intensity, IMIN, of the image 103 is obtained. If IMIN is not equal to zero, the minimum intensity value of the image 109 is adjusted to zero at a step 108. The re-adjustment procedure is carried out using Eq. 2:
Ib(x,y)=Iinvt(x,y)−IMIN (2)
Where Ib(x,y) (109) is a result of step 108. At step 110, Otsu's method is used to analyze the intensity histogram and calculate the optimal threshold of the micro-array image. By using this step, the optimal threshold level T (threshold 111) and its binary image It(x,y) (113) are obtained. If T is very low, a linear histogram transformation is used at step 112. The linear histogram transformation stretches the nonzero input intensity range:
At step 114, the optimal threshold image 115 is obtained from image 113. At Step 116 the re-sampled image 117 is obtained by decreasing the size of the image. The width and height of the image are repeatedly divided by 2 to decrease the size until the image width is less than a predefined the size. The ratio of re-sampling, r, is equal to 2n (where n is the number of times the image has been divided).
b) shows the pattern spectrum curve of the micro-array image. The image in
a) shows one sub-grid image 309 after the threshold operation.
where MEDx and MEDy are the mean values in the X and Y directions respectively, STDx and STDy are values of the standard deviation in the X and Y directions respectively, n is the height and m is the width of the image 409.
At step 504, the average values of STDx and STDy are calculated by:
where UX and UY are the average values of STDx and STDy respectively, m and n are the width and height of image 409.
At step 506, every STDx is compared with UX. If STDx is larger than UX, the position x is registered in an array Kx as a candidate of the spot centre position in the X direction. At step 508, similarly, every STDy is compared with UY, and the position y at which STDy is larger than UY is registered as a candidate of the spot centre position in the Y direction. At step 510, a spacing series in X or Y is obtained by calculation:
SPAx=Kx+1−Kx(x=0, 1, . . . , km) (10)
SPAy=Ky+1−Ky(y=0, 1, . . . , kn) (11)
where SPAx and SPAy are the spacing between the candidate of spot centres in the X and Y directions, and km and kn are the numbers of candidates for the spot positions in X and Y directions respectively.
At step 512, the values of the modes of SPAX and SPAy are calculated by the following equations:
LX=Mode(SPAx) (x=1, 2, . . . , px) (12)
LY=Mode(SPAy) (y=1, 2, . . . , py) (13)
where LX or LY are the values of the mode of SPAx or SPAy respectively, km and kn are the numbers of candidates of the spot positions in the X and Y directions respectively, and px and py are the numbers of spacing values between the spot candidates in the X and Y directions respectively. At step 514, Kx, Ky, LX, LY, and SP are used to obtain the spacing SDx and SDy between sub-grids and their positions Px and Py, sub-grid numbers GX and GY, and average numbers of spots in every sub-grid NX and NY in the X and Y directions. Every sub-grid area can be decided by Px and Py. At step 516, a typical sub-grid image 520 is made according to LX, LY, NX, NY and SP.
Numerous variations will be readily apparent to those skilled in the art within the scope of the attached claims.
Number | Name | Date | Kind |
---|---|---|---|
6349144 | Shams | Feb 2002 | B1 |
6577956 | Shams | Jun 2003 | B1 |
6674882 | Shams | Jan 2004 | B1 |
6990221 | Shams | Jan 2006 | B2 |
7110585 | Cork et al. | Sep 2006 | B2 |
7171030 | Foran et al. | Jan 2007 | B2 |
7189833 | Cohen et al. | Mar 2007 | B2 |
Number | Date | Country | |
---|---|---|---|
20050201602 A1 | Sep 2005 | US |