System for co-registration of medical images using a classifier
Disclosed is a system for analysis of microscopic image data representing a plurality of images acquired from cells. The system comprises a data processing system which is configured to read and/or generate ( 120 ) segmentation data for each of the images. For each of the images, the segmentation data are indicative of a segmentation of at least a portion of the respective image into one or more image regions so that each of the image regions is a member of one or more predefined classes of image content. The data processing system further generates co-registration data using at least portions of the segmentation data for co-registering at least portions of different ones of the images. The data processing system further generates mapping data using at least portions of the segmentation data for mapping between image regions of different images.
1 . A system for analysis of microscopic image data representing a plurality of images acquired from cells, the system comprising a data processing system which is configured to:
generate segmentation data for each of the images using a classifier of an artificial neural network, the classifier configured to perform sliding window classification, wherein for each of the images, the segmentation data are indicative of a segmentation of at least a portion of the respective image into three or more image regions so that each of the image regions is a member of a different one or more of a plurality of different predefined classes of image content, the plurality of different predefined classes of image content representing a plurality of different tissue types, wherein the plurality of different predefined classes comprise: (i) a class representing image regions formed by fatty tissue; (ii) a class representing image regions which are free from sample material; and (iii) a class representing image regions formed by non-fatty tissue;
generate co-registration data using at least portions of the segmentation data for co-registering at least portions of different images; and
generate mapping data using at least portions of the segmentation data for mapping between a plurality of image regions of different images, wherein generating mapping data comprises determining, for each of the image regions, an identification parameter for identifying the respective image region from among the remaining image regions contained in the same image, and wherein the identification parameter is determined depending on the segmentation data.
2 . The system of claim 1 , wherein a magnification of the segmentation data is lower than a magnification of the image data.
3 . The system of claim 1 , wherein the segmentation data comprise, for each of a plurality of pixels of the images, binary or probabilistic pixel classification data for providing a pixelwise classification of the pixels into one or more of the pre-defined classes.
4 . The system of claim 1 , wherein:
the classifier is based on supervised and/or unsupervised learning; and
the classifier is configured for performing at least a portion of a segmentation of the image data, wherein the segmentation generates the segmentation data using at least a portion of the image data.
5 . The system of claim 1 , wherein the image regions represent isolated tissue portions.
6 . The system of claim 1 , wherein the data processing system comprises a graphical user interface;
wherein the data processing system is configured to present to the user a one or more graphical representations which are generated depending on the co-registration data and/or depending on the mapping data.
7 . The system of claim 1 , wherein the system comprises an image acquisition unit which is configured to:
receive one or more samples, each of which comprising the cells; and
acquire the image data from the one or more samples.
8 . A method for analyzing microscopic image data representing a plurality of images acquired from cells using a data processing system, the method comprising:
generating, by the data processing system, segmentation data for each of the images using a classifier of an artificial neural network, the classifier configured to perform sliding window classification, wherein for each of the images, the segmentation data are indicative of a segmentation of at least a portion of the respective image into three or more image regions so that each of the image regions is a member of a different one or more of a plurality of different predefined classes of image content, the plurality of different predefined classes of image content representing a plurality of different tissue types, wherein the plurality of different predefined classes comprise two or more of: (i) a class representing image regions formed by fatty tissue; (ii) a class representing image regions which are free from sample material; and (iii) a class representing image regions formed by non-fatty tissue;
generating co-registration data using at least portions of the segmentation data for co-registering at least image portions of different images; and
generating mapping data using at least portions of the segmentation data for mapping between a plurality of image regions of different images, wherein generating mapping data comprises determining, for each of the image regions, an identification parameter for identifying the respective image region from among the remaining image regions contained in the same image, and wherein the identification parameter is determined depending on the segmentation data.
9 . The method of claim 8 , further comprising:
generating a first image of the images and a second image of the images, wherein the first image shows a sample being stained using a first stain and the second image shows a different and/or the same sample being stained using a second stain so that the first and second images show different sample stainings.
10 . A non-transitory computer readable medium having stored thereon a program element for analysis of microscopic image data representing a plurality of images acquired from cells, wherein the analysis is performed using a data processing system, wherein the program element, when being executed by a processor of the data processing system, is adapted to carry out:
generating, by the data processing system, segmentation data for each of the images using a classifier of an artificial neural network, the classifier configured to perform sliding window classification, wherein for each of the images, the segmentation data are indicative of a segmentation of at least a portion of the respective image into three or more image regions so that each of the image regions is a member of a different one or more of a plurality of different predefined classes of image content, the plurality of different predefined classes of image content representing a plurality of different tissue types, wherein the plurality of different predefined classes comprise two or more of: (i) a class representing image regions formed by fatty tissue; (ii) a class representing image regions which are free from sample material; and (iii) a class representing image regions formed by non-fatty tissue;
generating co-registration data using at least portions of the segmentation data for co-registering at least image portions of different images; and
generating mapping data using at least portions of the segmentation data for mapping between a plurality of image regions of different images, wherein generating mapping data comprises determining, for each of the image regions, an identification parameter for identifying the respective image region from among the remaining image regions contained in the same image, and wherein the identification parameter is determined depending on the segmentation data.
11 . The non-transitory computer readable medium of claim 10 , wherein a magnification of the segmentation data is lower than a magnification of the image data.
12 . The non-transitory computer readable medium of claim 10 , wherein the segmentation data comprise, for each of a plurality of pixels of the images, binary or probabilistic pixel classification data for providing a pixelwise classification of the pixels into one or more of the pre-defined classes.
13 . The non-transitory computer readable medium of claim 10 , wherein the program element is adapted to further carry out:
performing at least a portion of a segmentation of the image data, wherein the segmentation generates the segmentation data using at least a portion of the image data.
14 . The non-transitory computer readable medium of claim 10 , wherein the image regions represent isolated tissue portions.