Method for providing an aggregate algorithm for processing medical data and method for processing medical data
A method is for providing an aggregate algorithm for processing medical data. In an embodiment, a multitude of local algorithms are trained by machine learning. The training of each respective local algorithm is performed on a respective local system using respective local training data. A respective algorithm dataset concerning the respective local algorithm is transferred to an aggregating system that generates the aggregate algorithm based on the algorithm datasets.
1. A method for providing an aggregate algorithm for processing medical data, the method comprising:
training a plurality of local algorithms by machine learning by training each respective local algorithm, of the plurality of local algorithms, on a respective local system using respective local training data;
transferring a plurality of algorithm datasets corresponding to the plurality of local algorithms, respectively, to an aggregating system, each algorithm dataset of the plurality of algorithm datasets including at least one of a complete algorithm of the respective local algorithm or parameters for parameterizing a basic algorithm to reconstruct the respective local algorithm; and
generating the aggregate algorithm by combining outputs of the plurality of local algorithms by at least one of
a weighted or non-weighted averaging, or
majority voting.
2. The method of claim 1 , further comprising:
generating at least part of the local training data by
applying an initial algorithm to local input data to generate local output data,
providing a local data representation depending on the local output data to a user, and
storing at least one of a user feedback provided by the user or a user information determined from the user feedback as the at least part of the local training data.
3. The method of claim 2 , further comprising:
updating the aggregate algorithm in multiple iterations, wherein at least one iteration of the multiple iterations includes
generating a current iteration of the multiple iterations by using the aggregate algorithm determined during a previous iteration, of the multiple iterations, as the initial algorithm for the current iteration.
4. The method of claim 3 , further comprising:
training the initial algorithm by machine learning using initial training data.
5. The method of claim 3 , further comprising:
providing at least part of initial training data to the aggregating system; and
at least one of
validating the aggregate algorithm using the at least part of the initial training data, or
determining, based on the at least part of the initial training data, a respective quality measure for each local algorithm of the plurality of local algorithms and generating the aggregate algorithm based on the quality measures.
6. The method of claim 4 , further comprising:
providing at least part of the initial training data to the aggregating system; and
at least one of
validating the aggregate algorithm using the at least part of the initial training data, or
determining, based on the at least part of the initial training data, a respective quality measure for each local algorithm of the plurality of local algorithms and generating the aggregate algorithm based on the quality measures.
7. The method of claim 2 , further comprising:
training the initial algorithm by machine learning using initial training data.
8. The method of claim 7 , further comprising:
providing at least part of the initial training data to the aggregating system; and
at least one of
validating the aggregate algorithm using the at least part of the initial training data, or
determining, based on the at least part of the initial training data, a respective quality measure for each local algorithm of the plurality of local algorithms and generating the aggregate algorithm based on the quality measures.
9. The method of claim 8 , further comprising:
providing at least part of the initial training data to each of the local systems; and
training the plurality of local algorithms using the at least part of the initial training data.
10. The method of claim 7 , further comprising:
providing at least part of the initial training data to each of the local systems; and
training the plurality of local algorithms based on the at least part of the initial training data.
11. The method of claim 2 , wherein the user feedback includes at least one of
a modification of at least one of the local data representation or the local output data; or
a rating concerning a quality of at least one of the local data representation or the local output data.
12. The method of claim 1 , wherein at least one of the local algorithm, the aggregate algorithm or an initial algorithm is configured to
process image data as input data, and
output at least one of a segmentation of the image data or parameters concerning anatomy of a patient depicted by the image data.
13. A method for processing medical data, comprising:
processing the medical data by an aggregate algorithm, generated by the method of claim 1 , to generate output data.
14. A non-transitory computer program product storing a computer program, directly loadable into a memory of a processor, the computer program including instructions for performing the method of claim 10 when the program is executed by the processor.
15. A non-transitory computer-readable storage medium storing electronically readable instructions for performing the method of claim 10 when the electronically readable instructions are executed by a processor.
16. A non-transitory computer program product storing a computer program, directly loadable into a memory of a processor, the computer program including instructions for performing the method of claim 1 when the program is executed by the processor.
17. A non-transitory computer-readable storage medium storing electronically readable instructions for performing the method of claim 1 when the electronically readable instructions are executed by a processor.
18. The method of claim 1 , wherein the medical data is medical image data.
19. A system comprising:
at least one processor configured to cause the system to
train a plurality of local algorithms by machine learning by training each respective local algorithm, of the plurality of local algorithms, on a respective local system using respective local training data;
transfer a plurality of algorithm datasets corresponding to the plurality of local algorithms, respectively, to an aggregating system, each algorithm dataset of the plurality of algorithm datasets including at least one of a complete algorithm of the respective local algorithm or parameters for parameterizing a basic algorithm to reconstruct the respective local algorithm; and
generate the aggregate algorithm by combining outputs of the plurality of local algorithms by at least one of
a weighted or non-weighted averaging, or
majority voting.
20. The system of claim 19 , wherein the at least one processor comprises multiple further processors, each configured to perform a function of a local system.
21. The system of claim 19 , wherein the at least one processor is further configured to cause the system to:
generate output data by processing data by the aggregate algorithm.