METHOD FOR UNBIASED ESTIMATION OF THE TOTAL AMOUNT OF OBJECTS BASED ON NON UNIFORM SAMPLING WITH PROBABILITY OBTAINED BY USING IMAGE ANALYSIS

• US 20100142794A1
• Filed: 12/14/2007
• Published: 06/10/2010
• Est. Priority Date: 12/15/2006
• Status: Abandoned Application

• Current Assignee: Aarhus Universitet (Ministeriet For Forskning Innovation OG Videregående UD- Dannelser)
• Original Assignee: Aarhus Universitet (Ministeriet For Forskning Innovation OG Videregående UD- Dannelser)
• Inventors: Gardi, Jonathan Eyal, Gundersen, Hans Jørgen Gottlieb, Nyengaard, Jens Randel
• Application Number:
  US12/448,264
• Publication Number:
  US 20100142794A1
• Time in Patent Office: Days
• US Class Current: 382/133
• CPC Class Codes: G06T 2200/24 involving graphical user interfaces [GUIs]
• CPC Class Codes: G06T 2207/10056 Microscopic image
• CPC Class Codes: G06T 2207/20021 Dividing image into blocks, subimages or windows
• CPC Class Codes: G06T 2207/30024 Cell structures in vitro; Tissue sections in vitro
• CPC Class Codes: G06T 7/0012 Biomedical image inspection
• CPC Class Codes: G06T 7/90 Determination of colour characteristics
• CPC Class Codes: G06V 20/69 Microscopic objects, e.g. biological cells or cellular parts

Abstract

An image is partitioned into sectors, and a number of sectors are selected randomly but with a probability of selection which is proportional with the likelihood of objects in the sector. For the selected sectors, the objects are measured or counted and used for estimation of the amount of objects in the entire image.

Claims

1. Method for unbiased estimation of structural content, the method comprising the steps of

providing an image with discernible image analysis features indicating the objects,

by image analysis partitioning the image into a plurality of sectors and sampling a subset of the plurality of sectors, wherein the number of the sampled sectors is substantially less than the number of the plurality of sectors,

determining the structural content of objects for each of the sampled sectors, and

calculating an unbiased estimation for a total structural content of objects in the image based on the result from the determining of the structural content of objects in the sampled sectors,

the sampling of the subset of sectors is performed in accordance with a random sampling criterion

wherein

the sampling is performed in accordance with a random sampling criterion using a non-uniform probability which is positively related to the likelihood of object-presence in a sector

2. Method according to claim 1, wherein the method comprises

defining criteria for specific types of image analysis features, the image analysis features being indicative of the objects,

by computerised image analysis, automatically analysing the sectors with respect to the defined criteria and assigning a numerical weight factor z_i to each analysed sector, the weight factor being positively related, for example proportional, to structural content, for example total number or total amount, of detectable features,

sampling a number of sectors according to a random sampling criterion, wherein the probability for sampling of a specific sector is proportional to the weight factor z_i for the specific sector.

3. Method according to claim 2, wherein the random sampling criteria is the Systematic Uniform Random Sampling (SURS).

4. Method according to claim 3, further comprising calculating an accumulated weight Z for all sectors, selecting a sample size n, and setting the SURS period for sampling to Z/n.

5. Method according to claim 4, further comprising the step of arranging the sectors for sampling in accordance with the Smooth Fractionator based on the weights of the sectors, optionally, before calculating the accumulated weight Z.

6. Method according to claim 4, further comprising giving non-uniform sampling probability p_i to each sector, wherein p_i is equal to the weight factor z_i divided by the SURS period Z/n.

7. Method according to claim 6, further comprising sampling the subset of sectors by using SURS on the accumulated weight factors z_i in order to sample the sectors according to their probability.

8. Method according to claim 6, further comprising using the Horvitz-Thompson estimator with summing x_i/p_i for the subset of sectors for estimating the structural content, for example total number or the total amount, of objects.

9. Method according to claim 1, wherein the criteria for specific types of image analysis features is at least one from the group of colour criteria, morphology criteria and contextual criteria.

10. Method according to claim 9, wherein the colour criteria is represented by defined volume in a three dimensional colour space, the three dimensions in the colour space each given by numerical values for the saturation of red, green and blue.

11. Method according to claim 10, wherein the defined volume is a sphere.

12. Method according to claim 1, wherein the absolute precision in the form of a Coefficient of Errors is computed from two or more independent estimates based on samples of sectors of half or less the intended size using the standard error of the mean divided by the mean of these independent estimates.

13. Method according to claim 1, wherein an efficiency of the method relative to simple random sampling is computed automatically from the sample of sectors based on their known sampling probability zi and known content xi.

14. Method according to claim 1, wherein those sectors that are analysed automatically for image features positively related to the objects to be quantified are a statistically uniform subsample with a known probability of the large total number of sectors in the image.

15. Method according to claim 1, wherein the image is an aggregate of many images of a larger region.

16. Method according to claim 1, wherein the image is a two-dimensional map of information in an invisible part of the spectra of radiation.

17. Method according to claim 1, wherein the object is an anatomical structure.

18. Method according to claim 17, where the method implies determining the structural content of objects, for example counting or measuring the objects, by using stereology including the use of a physical dissector principle relying on two thin serial sections of a tissue sample or images thereof.

19. Method according to claim 1, wherein the image is a microscopy image.

20. Method according to claim 19, wherein the method comprises obtaining the microscopy image by use of a virtual microscope or a scanning microscope.

21. Method according to claim 19, wherein the method implies taking a microscopy image of a histological tissue section or of a cytological cell spread specimen.

22. Method according to claim 1, wherein the image is a satellite image of a geographical region or a telescopic image of a celestial region.