Histology recognition to automatically score and quantify cancer grades and individual user digital whole histological imaging device
Digital pathology is the concept of capturing digital images from glass microscope slides in order to record, visualize, analyze, manage, report, share and diagnose pathology specimens. The present disclosure is directed to a desktop slide scanner, which enables pathologists to scan slides at a touch of a button. Included is a workflow for reliable imaging, diagnosis, quantification, management, and sharing of a digital pathology library. Also disclosed herein is an analysis framework that provides for pattern recognition of biological samples represented as digital images to automatically quantitatively score normal cell parameters against disease state parameters. The framework provides a pathologist with an opportunity to see what the algorithm is scoring, and simply agree, or edit the result. This framework offers a new tool to enhance the precision of the current standard of care.
1. A computer-implemented method for determining and grading of features of a biological sample represented by a digital image, comprising:
performing an initial region classification to classify cells within the biological sample;
surveying a tumor region to assess disease state to perform a cancer cell classification;
grading the cancer cell classification of the biological sample; and
generating a report of the graded biological sample, wherein performing the initial region classification further comprises performing a Hematoxylin and Eosin (H&E) nucleus identification, and wherein the H&E nucleus identification comprises segmenting a nucleus by red-green-blue (RGB) values and selecting the nucleus in accordance with predetermined area and roundness criteria.
2. The method of claim 1 , performing the initial region classification further comprising:
applying a pattern recognition algorithm to the digital image to identify tumor cells.
3. The method of claim 2 , further comprising:
determining a number of tumor cells in the biological sample;
determining an 2D area of the tumor cells; and
determining a ratio of tumor cells to non-tumor cells in the biological sample.
4. The method claim 1 , further comprising performing a quality control checkpoint operation to classify outliers.
5. The method of claim 1 , further comprising:
classifying breast epithelial cells into a predetermined category; and
providing a notification that the biological sample has not been classified by the initial region classification.
6. The method of claim 1 , surveying the tumor region further comprising assessing nuclear pleomorphism.
7. The method of claim 6 , further comprising:
determining a nuclear parameter, the nuclear parameter being at least one of size, hematoxylin counterstain, shape, texture and nucleus to cytoplasmic ratio;
segmenting tumor cells by identifying nuclei using the nuclear parameter; and
comparing features of the identified nuclei to the same features of normal nuclei, wherein assessing nuclear pleomorphism further comprises generating a nuclear pleomorphism score based on the comparison.
8. The method of claim 7 , wherein the size is between 25 μm2 and 60 μm2, wherein the hematoxylin counterstain is between 160DR and 210DR, wherein the shape has a radius between 0.65 and 1.0 and a length between 0.45 and 1.0, wherein the texture Haralick value is between 0.7 and 0.9, and wherein the nucleus to cytoplasmic ratio is between 0.8 and 4.
9. The method of claim 1 , grading the cancer cell classification further comprising:
determining a nuclear waterfall of the cancer cell;
determining a mitotic density; and
determining region fractals.
10. The method of claim 9 , wherein determining a nuclear waterfall comprises:
loading nuclear identification and feature data; and
creating distribution plot data for predetermined nuclear features.
11. The method of claim 10 , wherein determining the mitotic density comprises scoring the distribution plot against a tissue of interest library.
12. The method of claim 11 , wherein determining regions fractals comprises:
identifying a library of fractal analysis of structure for the tissue of interest;
running a fractal dependent analysis.
13. The method of claim 1 , further comprising comparing the cancer cell classification with a standard scoring algorithm.
14. The method of claim 13 , wherein the standard scoring algorithm is the Nottingham Breast Cancer Score.
15. A computer-implemented method for determining and grading of features of a biological sample represented by a digital image, comprising:
performing an initial region classification to classify cells within the biological sample;
surveying a tumor region to assess disease state to perform a cancer cell classification;
grading the cancer cell classification of the biological sample; and
generating a report of the graded biological sample, wherein performing the initial region classification further comprises performing an Eosin cytoplasm identification, and wherein the Eosin cytoplasm identification comprises segmenting a nucleus by red-green-blue (RGB) values and determining a nucleus to cytoplasmic ratio.
16. The method of claim 15 , performing the initial region classification further comprising:
applying a pattern recognition algorithm to the digital image to identify tumor cells.
17. The method of claim 16 , further comprising:
determining a number of tumor cells in the biological sample;
determining an 2D area of the tumor cells; and
determining a ratio of tumor cells to non-tumor cells in the biological sample.
18. The method of claim 15 , further comprising performing a quality control checkpoint operation to classify outliers.
19. The method of claim 15 , further comprising:
classifying breast epithelial cells into a predetermined category; and
providing a notification that the biological sample has not been classified by the initial region classification.
20. The method of claim 15 , surveying the tumor region further comprising assessing nuclear pleomorphism.
21. The method of claim 20 , further comprising:
determining a nuclear parameter, the nuclear parameter being at least one of size, hematoxylin counterstain, shape, texture and nucleus to cytoplasmic ratio;
segmenting tumor cells by identifying nuclei using the nuclear parameter; and
comparing features of the identified nuclei to the same features of normal nuclei, wherein assessing nuclear pleomorphism further comprises generating a nuclear pleomorphism score based on the comparison.
22. The method of claim 21 , wherein the size is between 25 μm2 and 60 μm2, wherein the hematoxylin counterstain is between 160DR and 210DR, wherein the shape has a radius between 0.65 and 1.0 and a length between 0.45 and 1.0, wherein a texture Haralick value is between 0.7 and 0.9, and wherein the nucleus to cytoplasmic ratio is between 0.8 and 4.
23. The method of claim 15 , grading the cancer cell classification further comprising:
determining a nuclear waterfall of the cancer cell;
determining a mitotic density; and
determining region fractals.
24. The method of claim 23 , wherein determining a nuclear waterfall comprises:
loading nuclear identification and feature data; and
creating distribution plot data for predetermined nuclear features.
25. The method of claim 24 , wherein determining the mitotic density comprises scoring the distribution plot data against a tissue of interest library.
26. The method of claim 25 , wherein determining region fractals comprises:
identifying a library of fractal analysis of structure for the tissue of interest;
running a fractal dependent analysis.
27. The method of claim 15 , further comprising comparing the cancer cell classification with a standard scoring algorithm.
28. The method of claim 27 , wherein the standard scoring algorithm is the Nottingham Breast Cancer Score.
29. A computer-implemented method for determining and grading of features of a biological sample represented by a digital image, comprising:
performing an initial region classification to classify cells within the biological sample;
surveying a tumor region to assess disease state to perform a cancer cell classification;
grading the cancer cell classification of the biological sample; and
generating a report of the graded biological sample, wherein grading the cancer cell classification further comprises:
determining a nuclear waterfall of the cancer cell;
determining a mitotic density; and
determining region fractals.
30. The method of claim 29 , wherein determining a nuclear waterfall comprises:
loading nuclear identification and feature data; and
creating distribution plot data for predetermined nuclear features.
31. The method of claim 30 , wherein determining the mitotic density comprises scoring the distribution plot data against a tissue of interest library.
32. The method of claim 31 , wherein determining region fractals comprises:
identifying a library of fractal analysis of structure for the tissue of interest;
running a fractal dependent analysis.