Automatic image quality and feature classification method

1 . A method of assessing bioassay image quality, comprising:

performing a two-dimensional (2D) fast Fourier transform of a test image and a reference image to obtain spatial frequency domain transforms of the test image and the reference image;

obtaining coordinates for frequency rings in the spatial frequency domain transforms of the test image and the reference image;

calculating and normalizing a first weighted Fourier ring magnitude histogram for the frequency rings of the test image and a second weighted Fourier ring magnitude histogram for the frequency rings of the reference image;

computing a union of the first weighted Fourier ring magnitude histogram and the second weighted Fourier ring magnitude histogram and an intersection of the first weighted Fourier ring magnitude histogram and the second weighted Fourier ring magnitude histogram;

filtering the intersection and the union using a histogram bin cutoff to obtain a high-pass filtered intersection and a high-pass filtered union; and

outputting a metric indicating a degree of similarity between the test image and the reference image for a selected characteristic based on dividing a sum of all bins in the high-pass filtered intersection by a sum of all bins in the high-pass filtered union.

2 . The method of claim 1 , wherein the selected characteristic comprises image quality.

3 . The method of claim 1 , wherein test image and the reference image comprise images of cells in a bioassay.

4 . The method of claim 3 , wherein the selected characteristic comprises a cell count.

5 . The method of claim 3 , wherein the selected characteristic comprises a cell morphology.

6 . The method of claim 3 , wherein the selected characteristic comprises cell health, and wherein the reference image depicts one or more healthy cells.

7 . A system, comprising:

a computing device including a processor configured to execute instructions stored in non-transitory memory that, when executed, cause the processor to:

receive an image of a bioassay and a reference image for evaluating a selected characteristic of the image of the bioassay;

perform a two-dimensional (2D) fast Fourier transform of the image of the bioassay and the reference image to obtain spatial frequency domain transforms of the image of the bioassay and the reference image;

obtain coordinates for frequency rings in the spatial frequency domain transforms of the image of the bioassay and the reference image;

calculate and normalize a first weighted Fourier ring magnitude histogram for the frequency rings of the image of the bioassay and a second weighted Fourier ring magnitude histogram for the frequency rings of the reference image;

compute a union of the first weighted Fourier ring magnitude histogram and the second weighted Fourier ring magnitude histogram and an intersection of the first weighted Fourier ring magnitude histogram and the second weighted Fourier ring magnitude histogram;

filter the intersection and the union using a histogram bin cutoff to obtain a high-pass filtered intersection and a high-pass filtered union; and

output a degree of similarity between the image of the bioassay and the reference image for the selected characteristic based on dividing a sum of all bins in the high-pass filtered intersection by a sum of all bins in the high-pass filtered union.

8 . The system of claim 7 , further comprising an imager communicatively coupled to the computing device, and wherein the computing device receives the image of the bioassay from the imager.

9 . The system of claim 8 , wherein the non-transitory memory includes further instructions that, when executed by the processor, cause the processor to:

select the reference image from a plurality of candidate reference images based on the degree of similarity between each of the plurality of candidate reference images with respect to each other, wherein the reference image has a highest degree of similarity.

10 . The system of any one of claim 9 , wherein the selected characteristic comprises one or more of an image quality, a cell count, and a cell feature.

11 . The system of claim 9 , wherein the selected characteristic comprises a cell feature, and wherein the cell feature comprises at least one of a cell morphology, a cell size, cell health, and cell degradation.

12 . A method, comprising:

performing a two-dimensional (2D) fast Fourier transform of images of the bioassay and reference images to obtain spatial frequency domain transforms of the images of the bioassay and the reference images;

obtaining coordinates for frequency rings in the spatial frequency domain transforms of the images of the bioassay and the reference images;

calculating and normalizing a first weighted Fourier ring magnitude histogram for the frequency rings of each image of the bioassay and a second weighted Fourier ring magnitude histogram for the frequency rings of each reference image;

computing a union of the first weighted Fourier ring magnitude histogram and the second weighted Fourier ring magnitude histogram and an intersection of the first weighted Fourier ring magnitude histogram and the second weighted Fourier ring magnitude histogram;

filtering the intersection and the union using a histogram bin cutoff to obtain a high-pass filtered intersection and a high-pass filtered union;

determining a Fourier ring intersection over union (FRIOU) metric for the images of the bioassay relative to the reference images by dividing a sum of all bins in the high-pass filtered intersection by a sum of all bins in the high-pass filtered union; and

quantifying the feature of cells in the bioassay based on the FRIOU metric.

13 . The method of claim 12 , wherein the feature comprises a number of cells having the FRIOU metric greater than a pre-determined threshold.

14 . The method of claim 12 , wherein the feature comprises a relative change with respect to a treatment condition of the bioassay.

15 . The method of claim 12 , wherein the feature comprises a percentage of cells infected via an infectious agent used in the bioassay.

16 . The method of claim 12 , wherein the feature determined based on a highest FRIOU metric between the images of the bioassay and a reference image depicting the feature.

17 . The method of claim 12 , wherein the FRIOU metric indicates a degree of similarity for a given image of the bioassay relative to a given reference image.