US 11,989,960 B2
Detecting cells of interest in large image datasets using artificial intelligence
Dan Charles Wilkinson, Jr., Fremont, CA (US); and Benjamin Adam Burnett, New York, NY (US)
Assigned to BlueRock Therapeutics LP, Cambridge, MA (US)
Appl. No. 17/432,832
Filed by BlueRock Therapeutics LP, Cambridge, MA (US)
PCT Filed Feb. 20, 2020, PCT No. PCT/US2020/019109
§ 371(c)(1), (2) Date Aug. 20, 2021,
PCT Pub. No. WO2020/172460, PCT Pub. Date Aug. 27, 2020.
Claims priority of provisional application 62/808,054, filed on Feb. 20, 2019.
Prior Publication US 2022/0172496 A1, Jun. 2, 2022
Int. Cl. G06V 20/69 (2022.01); G06V 10/764 (2022.01); G06V 10/774 (2022.01); G06V 10/776 (2022.01); G06V 10/82 (2022.01)
CPC G06V 20/698 (2022.01) [G06V 10/764 (2022.01); G06V 10/774 (2022.01); G06V 10/776 (2022.01); G06V 10/82 (2022.01); G06V 20/695 (2022.01)] 8 Claims
OG exemplary drawing
 
1. A method for detecting one or more cells of interest in image datasets, the method comprising:
receiving a plurality of stained histological images each containing an independent channel;
binarizing pixel values of the independent channel in each of the plurality of stained histological images;
determining an area of interest in the binarized images by finding pixel areas in the independent channel that are connected and comprise an overall connected pixel area of a certain size, each area of interest defined by bounding coordinates;
cropping each area of interest based upon the bounding coordinates to generate a set of sub-images each comprising a cropped area of interest;
labeling each sub-image as positive or negative for a cell of interest;
selecting a final classification model for detecting the cell of interest in each sub-image; and
deploying the final classification model to detect cells of interest in unclassified image datasets,
wherein the binarizing pixel values of independent channel includes binarizing pixel values of at least one independent channel in images of a set of merged images, the set of merged images generated by merging at least the plurality of stained histological images that have a first independent channel applied with corresponding plurality of stained histological images that have a second independent channel applied.