System and method for providing an analytical result based on a medical data set using ml algorithms
AssigneeSiemens Healthcare GmbH
InventorsLutz DOMINICK, Vladyslav UKIS
Life of this patent
Estimating the term…
Abstract
Methods and systems for providing an analytical result by automated evaluation of a medical data set in a distributed runtime environment are provided. In response to a request to provide the analytical result from a client computing device, a suitable version of an ML algorithm is selected from an algorithm repository unit and applied in a back-end computing device. Selection is made on the basis of areas of application which, for each version in the algorithm repository unit, characterize the version.
Claims20
A computing device to provide an analytical result based on an evaluation of a medical data set, the computing device comprising:
- an interface linked, for data purposes, to a second computing device, the interface configured to
- receive medical data sets and a request to provide an analytical result of the medical data set from the second computing device, and
- provide the analytical result to the second computing device;
- an algorithm repository unit including
- at least one ML algorithm trained to provide an analytical result based on medical data sets, and
- for each ML algorithm, a plurality of areas of application, each area of application of an ML algorithm corresponding to a different type of preprocessing of the medical data set prior to application of the ML algorithm;
- and
- a processing unit configured to
- select an ML algorithm from the algorithm repository unit based on the request,
- select an area of application for the selected ML algorithm based on at least one of the request or the medical data set, and
- request preprocessing of the medical data set in accordance with a respective preprocessing associated with the selected area of application to provide a preprocessed data set, and
- apply the selected ML algorithm to the preprocessed medical data set to provide the analytical result.
A computer-implemented method for providing an algorithm repository unit, the algorithm repository unit having at least one ML algorithm and an area of application of the ML algorithm, the method comprising:
- providing an ML algorithm configured to provide an analytical result based on a medical data set;
- providing a medical training data set and an associated verified analytical result;
- applying preprocessing to the medical training data set to produce a preprocessed training data set;
- applying the ML algorithm to the preprocessed training data set to produce an intermediate analytical result;
- comparing the intermediate analytical result with the verified analytical result;
- adapting the ML algorithm based on the comparison;
- creating an area of application of the ML algorithm based on the comparison, the area of application including an indication of the applied preprocessing;
- and
- storing the adapted ML algorithm and the area of application in the algorithm repository unit.
A computer-implemented method for providing an analytical result by evaluating a medical data set with a computing device, the computer-implemented method comprising:
- receiving a request to provide an analytical result of the medical data set from a second computing device;
- providing, in the computing device, an algorithm repository unit having
- at least one ML algorithm trained to provide analytical results based on medical data sets, and
- for each ML algorithm, a plurality of areas of application, each area of application of an ML algorithm corresponding to a different type of preprocessing of the medical data set prior to application of the ML algorithm;
- selecting an area of application based on at least one of the request or the medical data set;
- requesting preprocessing corresponding to the preprocessing associated with the selected area of application to provide a preprocessed data set;
- and
- applying the ML algorithm to the preprocessed medical data set to provide the analytical result to the second computing unit.
Dependencies are inferred from the claim text.