Method for training a deep learning model to obtain histopathological information from images

1. A method for supervised training a deep learning model to obtain histopathological information from images, the method comprising the steps of:

providing a first image of a first section of a specimen, wherein the first section has been stained using a first staining protocol, wherein the staining of the first section is selected from the group of immunohistochemistry (IHC) and immunofluorescence (IF);

providing a second image of a second section of the specimen, the first and second sections being adjacent sections of the specimen, wherein the second section has been stained using a second staining protocol different from the first staining protocol;

co-registration of the first and second sections of the images, wherein a transfer map or transfer function expresses an alignment of the first section of the first image and the second section of the second image;

obtaining histopathological information for the first section of the first image, the step of obtaining histopathological information comprising automatic image analysis and automatic labelling and/or annotation of the first image based on the staining of the first section and the first staining protocol;

transferring the histopathological information for the first section of the first image to the second image based on the co-registration of the first and second sections, thereby automatically labelling and/or annotating the second image according to the histopathological information of the first image, the histopathological information comprising labelling and/or annotation of the first image, and according to the transfer map or transfer function; and

training a deep learning model, based on the automatically transferred labelling and/or annotation of the second image, to obtain histopathological information from images stained with the second staining protocol wherein the supervised training of the deep learning model comprises the step of comparing training annotations associated with the second image against the automatically transferred histopathological information.

2. The method according to claim 1 , wherein the deep learning model is trained to recognize expressions associated with the histopathological information of the first section, in images stained with the second staining protocol.

3. The method according to claim 1 , wherein the deep learning model is trained to recognize specific immunohistochemistry information in hematoxylin and eosin stained images.

4. The method according to claim 1 , wherein the first staining protocol is more specific and/or more reliable than the second staining protocol.

5. The method according to claim 1 , wherein the first staining protocol provides additional or different information than the second staining protocol.

6. The method according to claim 1 , wherein the first staining protocol is assumed to provide a true labelling.

7. The method according to claim 1 , wherein the labelled second images, in a further step, are used for training a deep learning model to obtain histopathological information from an image of a section of a specimen based on hematoxylin and eosin (H&E) staining, by recognizing specific immunohistochemistry (IHC)-based staining information.

8. The method according to claim 1 , wherein the labelled second images, in a further step, are used for training a deep learning model to obtain histopathological information from an image of a section of a specimen based on hematoxylin and eosin (H&E) staining, by recognizing specific immunofluorescence (IF)-based staining information.

9. The method according to claim 1 , wherein the first and second adjacent sections are obtained at a distance of less than 50 μm, or less than 20 μm, or less than 10 μm or less than 5 μm, or less than 3 μm.

10. The method according to claim 1 , wherein the first and second sections of the images are aligned on a cellular level using feature-based co-registration techniques.

11. The method according to claim 1 , wherein the first and second sections of the images are aligned using multi-scale point-matching and local deformations.

12. A system for training a deep learning model, comprising a computer-readable storage device for storing instructions that, when executed by a processor, performs the method for training a supervised deep learning model to obtain histopathological information from images according to claim 1 .