Predicting total nucleic acid yield and dissection boundaries for histology slides

Patent metadata

Field	Value
Publication number	US-11348240-B2
Application number	US-202017139798-A
Country	US
Kind code	B2
Filing date	Dec 31, 2020
Priority date	May 14, 2018
Publication date	May 31, 2022
Grant date	May 31, 2022

Abstract

A method for predicting an expected yield of nucleic acid from tumor cells within a dissection boundary on a hematoxylin and eosin (H&E) slide is provided.

First claim

What is claimed:

1. A computer-implemented method for predicting an expected yield of nucleic acid from tumor cells within a dissection boundary on a hematoxylin and eosin (H&E) slide, the method comprising:
receiving a digital image of the H&E slide at an image-based nucleic acid yield prediction system having one or more processors;
identifying, using the one or more processors, tumor cells of the H&E slide from the digital image using a trained cell segmentation model;
generating a tumor area mask based at least in part on the identified tumor cells, wherein the tumor area mask defines the dissection boundary of the H&E slide; and
predicting the expected yield of nucleic acid for the tumor cells within the dissection boundary by providing the digital image of the H&E slide to a model trained on a plurality of H&E slides having dissection labels and total nucleic yield labels.

2. The method of claim 1 , further comprising:
accepting an associated unstained slide of the H&E slide for next-generation sequencing when the predicted expected yield of nucleic acid exceeds a minimum threshold.

3. The method of claim 2 , wherein the minimum threshold is 50 ng.

4. The method of claim 1 , wherein when the predicted expected yield of nucleic acid fails to satisfy a target total nucleic acid yield:
identifying a number of associated unstained slides that satisfies the target total nucleic acid yield; and
accepting the number of associated unstained slides for next-generation sequencing.

5. The method of claim 4 , wherein the target total nucleic yield is selected from a range between and including 50 ng-2000 ng.

6. The method of claim 4 , wherein the associated unstained slides are flagged for scrapping.

7. The method of claim 4 , wherein the associated unstained slides comprise tissue from the same formalin-fixed paraffin embedded specimen.

8. The method of claim 4 , herein the number of associated unstained slides is estimated by dividing the target total nucleic yield by the predicted expected yield of the H&E slide and rounding up to the nearest integer.

9. The method of claim 8 , wherein sequencing is performed using the scrapings from the associated unstained slides.

10. The method of claim 4 , further comprising superimposing, using a viewer, the tumor area mask over the digital image of the associated unstained slides to visually indicate to a user which tumor cells to scrape.

11. The method of claim 4 , wherein generating the tumor area mask further comprises providing the digital image of the H&E slide to a model trained on a plurality of H&E slides having dissection labels.

12. The method of claim 1 , wherein a viewer superimposes the tumor area mask over the digital image of the associated unstained slides to visually indicate to a user which tumor cells to scrape.

13. The method of claim 1 , wherein generating the tumor area mask further comprises:
providing the digital image of the H&E slide to a model trained on a plurality of H&E slides having dissection labels.

14. The method of claim 1 , wherein predicting the expected yield of nucleic acid for the tumor cells within the dissection boundary further comprises:
providing the digital image of the H&E slide to a model trained on a plurality of H&E slides having dissection labels and total nucleic yield labels.

15. The method of claim 1 , wherein predicting the expected yield of nucleic acid for the tumor cells within the dissection boundary further comprises:
counting the number of tumor cells identified within the dissection boundary and multiplying the count by a tumor cell average nucleic acid yield.

16. The method of claim 1 , wherein predicting the expected yield of nucleic acid for the tumor cells within the dissection boundary further comprises:
calculating the surface area of the dissection boundary and multiplying the surface area by a dissection boundary average nucleic acid yield.

17. The method of claim 1 , wherein the dissection boundary is a microdissection boundary.

18. The method of claim 1 , wherein the dissection boundary is a macrodissection boundary.

19. The method of claim 1 , wherein the dissection boundary is a whole slide boundary.

20. The method of claim 1 , wherein identifying tumor cells of the H&E slide from the digital image using the trained cell segmentation model comprises:
applying, using the one or more processors, a plurality of tile images formed from the digital image to the trained cell segmentation model and, for each tile, assigning a cell classification to one or more pixels within the tile image.

21. The method of claim 20 , wherein assigning the cell classification to one or more pixels within the tile image comprises:
identifying, using the one or more processors, the one or more pixels as a cell interior, a cell border, or a cell exterior and classifying the one or more pixels as the cell interior, the cell border, or the cell exterior.

22. The method of claim 1 , wherein the trained cell segmentation model is a pixel-resolution three-dimensional UNet classification model trained to classify a cell interior, a cell border, and a cell exterior.

23. The method of claim 1 , wherein identifying tumor cells of the H&E slide from the digital image using the trained cell segmentation model comprises:
applying, using the one or more processors, each of a plurality of tile images formed from the digital image to the trained cell segmentation model; and
performing, using the one or more processors, a registration on segmented cells in each of the tile images by determining a cell border of each cell, determining a centroid of each cell, and shifting coordinates of centroids to a universal coordinate space for the digital image.

24. The method of claim 1 , wherein the trained cell segmentation model is trained using a set of H&E slide training images annotated with identified cell borders, identified cell interiors, and identified cell exteriors.

25. An image-based tumor cells prediction system configured to perform the method of claim 1 , the image-based tumor cells prediction system being contained within a pathology slide scanner system.

26. An image-based tumor cells prediction system configured to perform the method of claim 1 , the image-based tumor cells prediction system being contained partially within a pathology slide scanner system and partially within an external prediction computing system communicatively coupled to the pathology slide scanner system through a communication network.

27. The method of claim 1 , wherein the one or more processors are one or more graphics processing units (GPUs), tensor processing units (TPUs), and/or central processing units (CPUs).

Assignees

• Tempus Labs Inc

Inventors

• Yip Stephen
• Ho Irvin
• Sha Lingdao
• Osinski Boleslaw
• Khan Aly Azeem
• Kruger Andrew J
• Carlson Michael
• Greenwald Abel
• Willis Caleb
• Westley Andrew
• Jones Ryan
• Mahon Brett

Classifications

• G06T1/20 Primary

  Processor architectures; Processor configuration, e.g. pipelining · CPC title

• G06V10/44

  Local feature extraction by analysis of parts of the pattern, e.g. by detecting edges, contours, loops, corners, strokes or intersections; Connectivity analysis, e.g. of connected components · CPC title

• G06V10/764

  using classification, e.g. of video objects · CPC title

• G06T7/0012 Primary

  Biomedical image inspection · CPC title

• G06N3/045

  Combinations of networks · CPC title