System and method for analysis of microscopic image data and for generating an annotated data set for classifier training
Disclosed is a system for analysis of microscopic image data which includes a data processing system. Pixel classification data for each of a plurality of pixels of the microscopic image data are read. The pixel classification data include for each of the pixels of the microscopic image data, binary or probabilistic classification data for classifying the pixel of the microscopic image data into one or more object classes of pre-defined objects which are shown by the image. At least a portion of the pixels of the microscopic image data are grouped to form one or more pixels groups. For each of the pixel groups, probabilistic group classificati on data are calculated depending on the pixel classification data of the pixels of the respective group. The probabilistic group classification data are indicative of a probability that the group shows at least a portion of an object of the respective object class.
1. A system for analysis of microscopic image data, the system comprising a data processing system;
wherein the data processing system is configured to:
read or generate pixel classification data for each of a plurality of pixels of the microscopic image data, wherein the pixel classification data comprise, for each of the plurality of pixels of the microscopic image data, binary or probabilistic classification data for classifying the pixel into one or more object classes of predefined objects which are shown by the image;
group at least a portion of the pixels of the microscopic image data to form one or more pixels groups; and
calculate, for each of the pixel groups, probabilistic group classification data depending on at least a portion of the pixel classification data of the pixels of the respective group;
wherein for one or more of the object classes and for each of the pixel groups, the probabilistic group classification data are indicative of a probability that the respective group shows at least a portion of an object of the respective object class.
2. The system of claim 1 , wherein the data processing system comprises a user interface which is configured for interactive generation of an annotated data set for training a classifier using the probabilistic group classification data.
3. The system of claim 2 , wherein the interactive generation of the annotated data set comprises receiving user input which is indicative of a classification label for one of the pixel groups, wherein the classification label assigns the pixel group to one or more of the object classes or indicates that the pixel group does not represent an object of any of the predefined object classes.
4. The system of claim 3 , wherein the data processing system is configured to generate the pixel classification data using a classifier and to perform supervised training of the classifier using the pixels of at least a portion of the pixel groups.
5. The system of claim 2 , wherein the data processing system is configured to display, using the user interface of the data processing system, one or more of the pixel groups and for each of the displayed pixel groups a visually perceptible indicator which is generated depending on the probabilistic group classification data of the respective pixel group.
6. The system of claim 5 , wherein the indicator is indicative of an extent of the pixel group.
7. The system of claim 6 , wherein the data processing system is configured to generate the pixel classification data using a classifier and to perform supervised training of the classifier using the pixels of at least a portion of the pixel groups.
8. The system of claim 5 , wherein the data processing system is configured to generate the pixel classification data using a classifier and to perform supervised training of the classifier using the pixels of at least a portion of the pixel groups.
9. The system of claim 2 , wherein the data processing system is configured to generate the pixel classification data using a classifier and to perform supervised training of the classifier using the pixels of at least a portion of the pixel groups.
10. The system of claim 1 , wherein the one or more pixel groups are formed depending on at least a portion of the pixel classification data.
11. The system of claim 10 , wherein the one or more pixel groups are formed using a threshold value for the pixel classification data.
12. The system of claim 10 , wherein the data processing system is configured to generate the pixel classification data using a classifier and to perform supervised training of the classifier using the pixels of at least a portion of the pixel groups.
13. The system of claim 1 , wherein the data processing system is configured to generate the pixel classification data using a classifier and to perform supervised training of the classifier using the pixels of at least a portion of the pixel groups.
14. The system of claim 1 , wherein each of the pixel groups substantially represents a pixel cluster.
15. The system of claim 1 , wherein the data processing system is further configured to generate the pixel classification data depending on data generated using a classifier executed by the data processing system, wherein the data generated using the classifier comprise output data outputted by a layer of an artificial neural network of the classifier.
16. System of claim 15 , wherein the generation of the pixel classification data comprises applying a logit function to the output data and/or to data generated using the output data.
17. The system of claim 1 , wherein for each of the pixel groups, the corresponding probabilistic group classification data are determined depending on a mean value of the pixel classification data of at least a portion of the pixels of the respective pixel group.
18. A method for analysis of microscopic image data using a data processing system, the method comprising:
reading or generating pixel classification data for each of a plurality of pixels of the microscopic image data, wherein the pixel classification data comprise, for each of the plurality of pixels of the microscopic image data, binary or probabilistic classification data for classifying the pixel into one or more object classes of predefined objects which are shown by the image;
grouping at least a portion of the pixels of the microscopic image data to form one or more pixels groups; and
calculating, for each of the pixel groups, probabilistic group classification data depending on at least a portion of the pixel classification data of the pixels of the respective group;
wherein for one or more of the object classes and for each of the pixel groups, the probabilistic group classification data are indicative of a probability that the respective group shows at least a portion of an object of the respective object class.
19. A non-transitory computer readable medium comprising a program element encoded therein for analysis of microscopic image data 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:
reading or generating pixel classification data for each of a plurality of pixels of the microscopic image data, wherein the pixel classification data comprise, for each of the plurality of pixels of the microscopic image data, binary or probabilistic classification data for classifying the pixel into one or more object classes of predefined objects which are shown by the image;
grouping at least a portion of the pixels of the microscopic image data to form one or more pixels groups; and
calculating, for each of the pixel groups, probabilistic group classification data depending on at least a portion of the pixel classification data of the pixels of the respective group;
wherein for one or more of the object classes and for each of the pixel groups, the probabilistic group classification data are indicative of a probability that the respective group shows at least a portion of an object of the respective object class.