| CPC G06V 20/695 (2022.01) [G01N 1/30 (2013.01); G01N 1/34 (2013.01); G06T 7/0012 (2013.01); G06T 2207/20081 (2013.01); G06T 2207/20084 (2013.01); G06T 2207/30024 (2013.01)] | 23 Claims |
|
1. A method comprising:
obtaining target image data, the target image data representing a target tissue sample stained with hematoxylin and eosin; and
applying the target image data to a trained neural network pattern analyser to identify regions of interest that exhibit a specified disease state within the target image data, wherein the trained neural network has been trained by:
receiving first image data, the first image data representing a training tissue sample stained with a first staining agent, wherein the first staining agent comprises hematoxylin and eosin;
receiving second image data, the second image data representing the training tissue sample after washing the training tissue sample to remove the first staining agent from the training tissue sample and staining the training tissue sample with a second, different staining agent, wherein the second staining agent comprises an immunohistochemical (IHC) stain that stains cells in the training tissue sample that exhibit the specified disease state;
registering the first and second image data to generate registration data;
receiving data indicating a region of interest represented in the second image data that depicts cells exhibiting the specified disease state;
determining a mask in the first image data based upon the region of interest represented in the second image data and the registration data; and
based on the first image data and the mask in the first image data, training the neural network pattern analyser to identify regions of interest that exhibit the specified disease state within image data that represents tissue samples stained with hematoxylin and eosin.
|