CPC G06T 7/0012 (2013.01) [G06V 20/695 (2022.01); G06T 2207/10004 (2013.01); G06T 2207/30024 (2013.01); G06T 2207/30242 (2013.01)] | 20 Claims |
1. A computer-implemented method comprising:
accessing an image of a stained biological sample;
processing the image to generate predictive cell motility labels for the image, wherein the image is processed by:
detecting, using a trained classifier, lymphocytes in the image;
identifying outlines of the lymphocytes by applying image enhancing filters to the image, wherein applying the image enhancing filters to the image causes one or more regions of the image to be discarded and one or more regions of the image to be retained;
computing shape metrics for the lymphocytes based on the outlines, wherein a shape metric of the shape metrics is computed for each lymphocyte of the lymphocytes;
for each respective lymphocyte of the lymphocytes, assigning a respective predictive cell motility label of a plurality of predictive cell motility labels to the respective lymphocyte based on a respective shape metric of the shape metrics, the respective shape metric indicative of a shape of the respective lymphocyte, wherein the respective predictive cell motility label corresponds to a letter and/or number that identifies a likelihood that the respective lymphocyte is motile; and
generating an overlay for the image using the predictive cell motility labels, wherein the overlay provides a visualization of the lymphocytes and predicted motilities of the lymphocytes.
|