Transformation of histochemically stained images into synthetic immunohistochemistry (IHC) images
Inventors
Wang, Xingwei • Lorsakul, Auranuch • Zhao, Zuo • Nie, Yao • Koeppen, Hartmut • KOLSTER, Hauke • Shanmugam, Kandavel
Assignees
Interested in licensing this patent?
MTEC can help explore whether this patent might be available for licensing for your application.
Publication Number
US-12675880-B2
Publication Date
2026-07-07
Expiration Date
Abstract
The present disclosure relates to techniques for obtaining a synthetic immunohistochemistry (IHC) image from a histochemically stained image. Particularly, aspects of the present disclosure are directed to accessing an input image that depicts a tissue section that has been stained with at least one histochemical stain; generating a synthetic image by processing the input image using a trained generator network; and outputting the synthetic image. The synthetic image depicts a tissue section that has been stained with at least one IHC stain that targets a first antigen, and techniques may also include receiving an input that is based on a level of expression of a first antigen from the synthetic image and/or generating, from the synthetic image, a value that is based on a level of expression of the first antigen.
Core Innovation
The invention provides digital pathology image transformation that converts an input image depicting a tissue section stained with at least one histochemical stain into a synthetic image depicting the same tissue section as stained with at least one immunohistochemistry (IHC) stain. The synthetic image is generated by processing the input image using a generator network trained using a training data set that includes a plurality of pairs of images corresponding to the tissue section, where each pair includes a histochemically stained image and a corresponding IHC-stained image.
For each pair of images in the plurality of pairs, the histochemically stained image of a first section of a tissue and the IHC-stained image of a second section of the tissue are registered with each other. The trained generator network outputs the synthetic image for further processing.
The synthetic image is used to generate a value derived from the synthetic image representing a level of expression of a first antigen, and to identify a data point based on the level of expression of the first antigen, where identifying the data point includes receiving an input corresponding to the data point and/or generating the data point. The synthetic image depicts a tissue section stained with at least one IHC stain that targets the first antigen.
Claims Coverage
The document includes three independent claims, covering a method, system, and computer-program product, with five inventive features.
Synthetic image generation from histochemical stained tissue
Accessing an input image depicting a tissue section stained with at least one histochemical stain and generating a synthetic image by processing the input image using a generator network.
Training on paired registered histochemical and IHC images
Training the generator network using a training data set that includes a plurality of pairs of images corresponding to the tissue section, each pair including a histochemically stained image and an IHC-stained image, wherein the images are registered with each other.
Antigen expression value derived from synthetic image
Generating, from the synthetic image, a value derived from the synthetic image that represents a level of expression of a first antigen.
Data point identification based on antigen expression level
Identifying a data point based on the level of expression of the first antigen, where identifying the data point includes receiving an input corresponding to the data point and/or generating the data point.
Synthetic image depicting IHC stain targeting the first antigen
The synthetic image depicts a tissue section stained with at least one IHC stain that targets the first antigen.
Across the independent claims, the method, system, and computer-program product generate a synthetic IHC image from a histochemical input using a generator network trained on registered paired histochemical/IHC images, and then use the synthetic image to derive an antigen expression value and/or identify a data point based on the level of expression of a first antigen.
Stated Advantages
Reduced cost and time of acquiring IHC slides.
Enabling large synthetic datasets for AI verification and/or training.
Documented Applications
Supporting diagnosis determination based on antigen expression derived from the synthetic image and corresponding data points.
Supporting administering treatment decisions based on diagnosis-related outcomes derived from antigen expression.
Interested in licensing this patent?