US 12,450,927 B2
Histological stain pattern and artifacts classification using few-shot learning
Christian Roessler, Tucson, AZ (US); Yao Nie, Tuscon, AZ (US); Nazim Shaikh, Tucson, AZ (US); and Daniel Bauer, Tucson, AZ (US)
Assigned to Ventana Medical Systems, Inc., Tucson, AZ (US)
Filed by Ventana Medical Systems, Inc., Tucson, AZ (US)
Filed on Feb. 15, 2023, as Appl. No. 18/169,641.
Application 18/169,641 is a continuation of application No. PCT/US2021/046242, filed on Aug. 17, 2021.
Claims priority of provisional application 63/069,421, filed on Aug. 24, 2020.
Prior Publication US 2023/0196803 A1, Jun. 22, 2023
Int. Cl. G06K 9/00 (2022.01); G06V 10/74 (2022.01); G06V 10/764 (2022.01); G06V 10/771 (2022.01); G06V 10/82 (2022.01); G06V 20/69 (2022.01)
CPC G06V 20/698 (2022.01) [G06V 10/761 (2022.01); G06V 10/764 (2022.01); G06V 10/771 (2022.01); G06V 10/82 (2022.01)] 8 Claims
OG exemplary drawing
 
1. A computer-implemented method comprising:
accessing a pre-trained machine-learning model, the pre-trained machine-learning model including a set of parameter values corresponding to a set of parameters, wherein the set of parameter values were learned using a first training data set, the first training data set including a plurality of images and a corresponding set of classifications, each classification of the set of classifications characterizing a depiction in a corresponding image of the plurality of images;
accessing a set of digital pathology images and a corresponding set of artifact classifications, wherein each digital pathology image of the set of digital pathology images depicts a stained section of a sample and further includes an artifact, and wherein each artifact classification of the set of artifact classifications indicates a type of artifact corresponding to the artifact, wherein the set of classifications of the first training data set is different than the set of artifact classifications;
using few-shot learning to further train the pre-trained machine-learning model using the set of digital pathology images and the corresponding set of artifact classifications, wherein the further training generates a new set of parameter values for the set of parameters;
receiving a new digital pathology image;
processing the new digital pathology image using the further trained machine-learning model to generate an output predicting that the new digital pathology image includes a particular type of artifact;
determining a portion of the new digital pathology image that depicts the particular type of artifact; and
excluding the portion of the new digital pathology image from a subsequent digital pathology analysis.