Method, system, and computer code for finding spots defined in biological microarrays

• US 6,980,677 B2
• Filed: 05/24/2002
• Issued: 12/27/2005
• Est. Priority Date: 05/24/2002
• Status: Expired due to Fees

• Current Assignee: Niles Scientific, Inc.
• Original Assignee: Niles Scientific, Inc.
• Inventors: Foskett, Darach B., Martens, Christine L., Niles, Richard K.
• Primary Examiner(s): Patel, Kanjibhai
• Assistant Examiner(s): Tabatabai, Abolfazl
• Application Number:
  US10/155,781
• Publication Number:
  US 20030219150A1
• Time in Patent Office: 1,313 Days
• Field of Search: 382/128, 382/12, 382/130, 382/131, 382/132, 382/133, 382/134, 382/254, 382/261, 382/262, 382/263, 382/264, 382/265, 382/266, 382/322, 382/129, 435/6, 435/9, 435/30, 435/32, 435/34, 435/7.1, 435/7.2, 422/50, 422/52, 422/56, 422/57, 424/50, 424/52, 424/56, 436/63, 436/74, 436/519, 607/35
• US Class Current: 382/128
• CPC Class Codes: G06F 18/00 Pattern recognition
• CPC Class Codes: G06T 2207/30072 Microarray; Biochip, DNA array; Well plate
• CPC Class Codes: G06T 7/74 involving reference images or patches
• CPC Class Codes: G06V 2201/04 Recognition of patterns in DNA microarrays

Abstract

A method (and system) for using information contained within the scanned image to create, in an automated (or semi-automated) process, an accurate data grid. The process has steps: enhance the image; locate blocks of spots; and find each individual spot in each of the blocks. Preferably, the method makes use of image filtering using a “Principal Frequency Filter” based on a mathematical determination of major periodic elements in the image to eliminate noisy, non-periodic signals, and of smoothed intensity profiles of the filtered image data. Here, the term Principal Frequency Filter is used to indicate an image-enhancing filter based upon a mathematical operation which identifies the major periodic components of the image.

Claims

1. A method for processing data in digitized images of biological microarrays to identify one or more groupings of spots present in the microarray, the method comprising:

importing a captured digitized image of a biological microarray of spots in an electronic format, the array comprising a plurality of approximately rectangular groupings of spots, called blocks, the groupings being defined by 1 through N, where N is an integer greater than 1, the blocks being arranged in a regular pattern, with rows and columns of blocks being separated by substantially horizontal and vertical isolation regions comprising background, the background regions being long, narrow areas approximating a rectangular shape, which are substantially free from any spots;

processing the captured image to reduce background noise from the captured image;

identifying at least the isolation region between the groupings in the captured image using a frequency domain filter applied to the captured image, the filter being constructed according to periodic components of the captured image, the periodic components being defined by a spatial distribution of the spots in the captured image of the microarray;

determining the locations of the boundaries of the groupings in the captured image of the biological microarray to isolate any one of the groupings from any one of the other groupings; and

storing the locations of the boundaries of the groupings in the captured image of the biological microarray into memory.

2. The method of claim 1 wherein the filter is derived from a one- or multi-dimensional power spectrum from a Fourier transform process.

3. The method of claim 1 wherein the isolation regions are identified along a first dimension axis of the image and subsequently along the second dimension axis of the image.

4. The method of claim 1 wherein the locations of the boundaries of the blocks are determined from the filtered image by computing intensity profiles, the intensity profiles being calculated by summing pixel values along rows or columns, smoothing said intensity profile, and identifying a threshold value which the smoothed intensity profile crosses at locations consistent with estimated block size.

5. The method of claim 1 further comprising locating each of the spots in the isolated block using a filter applied to the captured image of the isolated block, the filter being defined according to periodic components of the captured image of the isolated block.

6. The method of claim 1 wherein the means for executing the method of claim 4 is provided in one or more computer systems.

7. A method for processing data in a captured image of a biological microarray to identify the location of one or more spots defined in the microarray image, the method comprising:

importing a captured digitized image of a biological microarray of spots in an electronic format, the array comprising a plurality of spots on a substrate material, the spots being defined by 1 through N, where N is an integer greater than 1, the spots being approximately equal in size and shape, the spots having been placed on the microarray in a predefined pattern which approximates a two dimensional lattice, each of the spots being identified by its row and column location in said lattice, each of the spots being provided on a background region, the spots generally distinguishable from the background region in the image by having higher values than the background in the case of a positive image or lower values than the background in the case of a negative image;

enhancing the electronic image to improve the visibility of the spots against the background region;

identifying the locations of the spots in the captured image by searching the enhanced electronic image for its local peak values;

relating the identified location of each spot to its row and column location in the two dimensional grid by predicting each spot'"'"'s location from information about said grid and matching identified locations with predicted locations; and

relating the one or more spots to respective one or more identifier information whereupon the identifier information uniquely identifies the spot.

8. The method of claim 7 wherein the identifier information is characterized as location information for the spot on the array.

9. The method of claim 8 wherein the identifier information is associated with process information for the spot, the process information being provided in one or more memories.

10. The method of claim 7 wherein the image enhancement is achieved by applying a filter to the captured image of the biological microarray, the filter being defined according to a periodic component of the spots in the captured image of the biological microarray.

11. The method of claim 10 wherein the filter is a principal frequency filter, said filter being constructed by selecting two of the strongest spatial-frequency components and eliminating other components of the captured image.

12. The method of claim 11 wherein the principal frequency filter is derived from a one-dimensional power spectrum or multidimensional power spectrum of a Fourier transform process.

13. The method of claim 7 wherein the relating of spots to identifier information is derived by means of a predicted grid, said grid being used as a map to predict locations of spots and to relate predicted locations to actual locations.

14. The method of claim 13 wherein the predicted grid is constructed from the two peak frequency components identified in the principal frequency filter by a matrix inversion calculation.

15. The method of claim 13 wherein predicted grid locations are related to actual spot locations by assigning each actual spot location to its nearest predicted grid location.

16. The method of claim 15 wherein a predicted spot location is related to its actual spot location by identifying the nearest neighboring location in the predicted grid to which a spot has already been assigned, and assigning the spot to the next location.

17. The method of claim 16 wherein the assignment of actual spot locations starts in the center of the image and proceeds outward from the center in a growing spiral.

18. A system for processing data in captured digitized images of biological microarrays to identify the location of one or more spots defined in the microarray image, the system comprising one or more memories, the one or more memories including:

a code directed to receiving a captured image of a biological microarray of spots in an electronic format, the array comprising a plurality of spots on a substrate material, the spots being defined by 1 through N, where N is an integer greater than 1, the spots being approximately equal in size and shape, the spots having been placed on the microarray in a predefined pattern which approximates a two dimensional lattice, each of the spots being identified by its row and column location in said lattice, each of the spots being provided on a background region, the spots generally distinguishable from the background region in the image by having higher values than the background in the case of a positive image or lower values than the background in the case of a negative image;

a code directed to enhancing the electronic image to improve the visibility of the spots against the background region;

a code directed to identifying the locations of the spots in the captured image by searching the enhanced electronic image for its local peak values;

a code directed to relating the identified location of each spot to its row and column location in the two dimensional grid by predicting each spot'"'"'s location from information about said grid and matching identified locations with predicted locations; and

a code directed to relating the one or more spots to respective one or more identifier information whereupon the identifier information uniquely identifies the spot.

19. The method of claim 18 wherein the image enhancing code is directed to calculate the principal frequency of one or more periodic components of the image, that principal frequency being used to construct the filter.

20. The method of claim 19 wherein the principal frequency filter code is directed to derive the filter from a one-dimensional power spectrum or multidimensional power spectrum of a Fourier transform process.

21. The method of claim 18 wherein the code directed to relate the one or more spots to respective one or more identifier information operates by means of a predicted grid, said grid being used as a map to predict locations of spots and to relate predicted locations to actual locations.

22. The method of claim 21 wherein the code directed to construct the predicted grid derives grid characteristics from the two-dimensional frequency and phase information of the two peak frequency components identified in the principal frequency filter.

23. The method of claim 21 wherein the code directed to construct the predicted grid derives grid characteristics from a cross correlation of the original image with itself.

24. The method of claim 21 wherein the code directed to relate predicted grid locations to actual spot locations operates by assigning each actual spot location to its nearest predicted grid location.

25. The method of claim 24 wherein the code directed to relate a predicted spot location to an actual spot location operates by identifying the nearest neighboring location in the predicted grid to which a spot has already been assigned, and assigning the spot to the next location.

26. The method of claim 24 wherein the code directed to assign actual spot locations operates by starting in the center of the image and proceeding outward from the center in a growing spiral.

27. A method for processing data in biological microarrays to identify one or more spots defined in the microarray, the method comprising:

importing a captured digitized image of a biological microarray of spots in an electronic format, the array comprising a plurality of spots on a substrate material, the spots being defined by 1 through N, where N is an integer greater than 1, the spots being approximately equal in size and shape, the spots having been placed on the microarray in a pattern which approximates a two dimensional grid, each of the spots being identified by its row and column location in said grid, each of the spots being provided on a background region, the background region including a covered region occupied by spots and an uncovered region being substantially free from any spots;

mathematically defining a filter based on the principal frequency of the repeating spot grid;

applying the filter to the captured image of the biological microarray to derive an enhanced image of the captured image, the filter being applied to enhance a spatial characteristic of the captured image;

processing the enhanced image to identify spatial positions of the spots in the captured image; and

relating the spatial positions of the spots to respective identifier information whereupon the identifier information uniquely identifies each of the spots.

28. The method of claim 27 wherein the filter is a principal frequency filter which identifies two or more spatial frequency components of an intensity higher than other frequency components that form the captured image.

29. The method of claim 28 wherein the step of defining the principal frequency filter is provided using a Fourier transform process.

30. The method of claim 27 wherein the relating of spots to identifier information is derived by means of a predicted grid, said grid being used as a map to predict locations of spots and to relate predicted locations to actual locations.

31. The method of claim 30 wherein the predicted grid is constructed from the two-dimensional frequency and phase information from the two peak frequency components identified in the principal frequency filter.

32. The method of claim 30 wherein the predicted grid is constructed from a cross correlation of the original image with itself.

33. The method of claim 30 wherein predicted grid locations are related to actual spot locations by assigning each actual spot location to its nearest predicted grid location.

34. The method of claim 33 wherein a predicted spot location is related to its actual spot location by identifying the nearest neighboring location in the predicted grid to which a spot has already been assigned, and assigning the spot to the next location.

35. The method of claim 34 wherein the assignment of actual spot locations starts in the center of the image and proceeds outward from the center in a growing spiral.

36. The method of claim 27 in the processing comprises locating a local peak of intensity associated with a portion of the captured image, the local peak of intensity being associated with one of the spots in the captured image.

37. The method of claim 27 wherein the filter is a frequency domain filter derived from the two dimensional power spectrum of the captured image.

38. The method of claim 27 wherein the identifier information comprises at least location information for the spot on the array.

39. A system for processing data in biological microarrays to identify one or more spots defined in the microarray, the system comprising one or more memories, the one or more memories including:

a code directed to receiving an image of a biological microarray of spots in an electronic format, the array comprising a plurality of spots on a substrate material, the spots being defined by 1 through N, where N is an integer greater than 1, the spots having been placed on the substrate material in a pattern approximating a two dimensional grid, each of the spots being identified by its row and column location in said grid, each of the spots being provided on a background region, the background region including a covered region occupied by spots and an uncovered region being substantially free from any spots;

a code directed to applying a filter to the captured image of the biological microarray to derive an enhanced image of the captured image, the filter being applied to enhance a spatial characteristic of the captured image;

a code directed to identifying the locations of the spots by searching the enhanced electronic image for its highest values;

a code directed to constructing a predicted spot location grid from frequency and phase information from the enhanced spatial characteristics of the captured image; and

a code directed to processing the enhanced image to identify spatial positions of the spots in the captured image.

40. The system of claim 39 further comprising a code directed to relating the spatial positions of the spots in the captured image to respective identifier information whereupon the identifier information uniquely identifies each of the spots.

41. The system of claim 40 comprising a code directed to relating the spatial positions of spots in the captured image by means of a predicted grid, said grid being used as a map to predict locations of spots and to relate predicted locations to actual locations.

42. The system of claim 41 comprising a code directed to constructing the predicted grid by deriving grid coordinates from the two-dimensional frequency and phase information of the two principal frequency components.

43. The system of claim 41 comprising a code directed to constructing the predicted grid by deriving grid coordinates from a cross-correlation of the captured image with itself.

44. The system of claim 40 comprising a code directed to relating predicted grid locations to actual spot locations by assigning each actual spot location to its nearest predicted grid location.

45. The system of claim 40 comprising a code directed to relating a predicted grid spot location to actual spot location by identifying the nearest neighboring location in the predicted grid to which a spot has already been assigned, and assigning the spot to the next location.

46. The system of claim 40 comprising a code directed to assigning actual spot locations by starting in the center of the image and proceeding outward from the center in a growing spiral.

47. The system of claim 39 wherein the filter is a principal frequency filter defined by identification of two or more spatial frequency components of higher intensity than other frequency components that form the captured image, and eliminating other components of the captured image, the principal frequency filter being provided using a multidimensional power spectrum from a Fourier transform process.

48. A method for processing data in biological microarrays to identify one or more spots defined in the microarray, the method comprising:

providing an image of a biological microarray of spots in an electronic format, the image being a captured image, the array comprising a plurality of spots on a substrate material, the spots being defined by 1 through N, where N is an integer greater than 1, the spots having been placed on the microarray in a pattern approximating a two dimensional grid, each of the spots being identified by its row and column location in said grid, each of the spots being provided on a background region, the background region including a covered region occupied by spots and an uncovered region being substantially free from any spots;

processing a two dimensional representation associated with the captured image of the biological microarray using a processing filter to provide an enhanced image of the captured image, the filter being derived from a two dimensional power spectrum of the two dimensional representation associated with the captured image by selecting two of the strongest non-zero frequency components and removing other frequency components outside of a frequency domain region associated with the two of the strongest non-zero frequency components;

processing the enhanced image to identify spatial positions of the spots in the captured image; and

relating the spatial positions of the spots to respective identifier information whereupon the identifier information uniquely identifies each of the spots.

49. The method of claim 48 wherein the frequency domain region is a frequency domain neighborhood of two of the strongest non-zero frequency components.