Dot detection, color classification of dots and counting of color classified dots
AssigneeVentana Medical Systems, Inc.
InventorsAnindya Sarkar, Jim Martin
Life of this patent
Estimating the term…
Abstract
The present disclosures relates to a method of detecting, classifying, and counting dots in an image of a tissue specimen comprising detecting dots in an image of the tissue sample that meet criteria for absorbance strength, black unmixed image channel strength, red unmixed image channel strength, and a difference of Gaussian threshold, wherein the detected dots correspond to in situ hybridization signals in the tissue samples; classifying the detected dots as belonging to a black in situ hybridization signal or to a red in situ hybridization signal; and calculating a ratio of those dots belonging to the black in situ hybridization signal and those belonging to the red in situ hybridization signal.
Claims22
A system for detecting first and second dots representing in-situ hybridization signals having different colors, the first and second dots being contained in a digital image of a tissue sample, the pixel intensity values of the first dots indicating in situ hybridization signals generated by a first stain having a first color, the pixel intensity values of the second dots indicating in situ hybridization signals generated by a second stain having a second color, the system comprising:
- (i) one or more processors, and (ii) a memory coupled to the one or more processors, the memory to store computer-executable instructions that, when executed by the one or more processors, cause the one or more processors to perform operations comprising:
- generating a first color channel image and a second color channel image via color deconvolution of the digital image, the first color channel image corresponding to the color spectrum contribution of the first stain and the second color channel image corresponding to the color spectrum contribution of the second stain;
- computing at least one DoG image from the digital image by applying a pair of Gaussian filters whose Kernels have a different standard deviation to the digital image and by subtracting the two filtered images output by the Gaussian filters from each other, the DoG image being a difference-of-Gaussian image;
- computing an absorbance image from the image of the tissue sample;
- detecting sets of adjacent pixels in the digital image whose absorbance values in the absorbance image exceed an absorbance threshold and whose DoG values in the DoG image exceeds an DoG threshold, and using the detected sets of adjacent pixels as prospective dots;
- identifying the ones of the prospective dots whose intensity values in the first color channel image exceeds a first-color-intensity threshold and outputting said identified prospective dots as the detected first dots;
- and
- identifying the ones of the prospective dots whose intensity values in the second color channel image exceeds a second-color-intensity threshold and outputting said identified prospective dots as the detected second dots.
A non-transitory computer-readable medium for storing computer-executable instructions that are executed by a processor to perform operations comprising:
- detecting dots in an image of the tissue sample that meet criteria for absorbance strength, black unmixed image channel strength, red unmixed image channel strength, and a difference of Gaussian threshold, wherein the detected dots correspond to in situ hybridization signals in the tissue samples, and wherein the dots that meet criteria for the absorbance strength, the black unmixed image channel strength, the red unmixed image channel strength, and the difference of Gaussian threshold are detected by computing a union of dot pixels and black/red pixels meeting predefined threshold criteria.
- classifying the detected dots as belonging to a black in situ hybridization signal corresponding to HER2 or to a red in situ hybridization signal corresponding to Chromosome 17;
- and
- calculating a ratio of those classified dots belonging to the black in situ hybridization and those belonging to the red in situ hybridization signal.
A computer-implemented method of detecting, classifying, and counting dots in an image of a tissue specimen comprising:
- detecting dots in an image of the tissue sample that meet criteria for absorbance strength, black unmixed image channel strength, red unmixed image channel strength, and a difference of Gaussian threshold, wherein the detected dots correspond to in situ hybridization signals in the tissue samples;
- classifying the detected dots as belonging to a black in situ hybridization signal corresponding to HER2 or to a red in situ hybridization signal corresponding to Chromosome 17;
- and
- calculating a ratio of those dots belonging to the black in situ hybridization signal and those belonging to the red in situ hybridization signal, wherein the ratio is calculated based on a count of red dot blobs and black dot blobs, wherein the red dot blobs are identified by connected components labeling of all classified dots belonging to the red in situ hybridization signal and wherein the black dot blobs are identified by connected components labeling of all classified dots belonging to the black in situ hybridization signal.
Dependencies are inferred from the claim text.