Method and arrangement for identifying similar pre-stored medical datasets

1. A method for identifying similar pre-stored medical datasets for comparison with a current case dataset, the method comprising:

providing a current case dataset comprising radiological data of a patient;

providing a number of pre-stored medical datasets each comprising radiological data of other patients;

obtaining features for the current case dataset and each of the number of pre-stored medical datasets based on an evaluation of the current case dataset and the number of pre-stored medical datasets according to a predefined AI-based method comprising a convolutional neural network configured to evaluate datasets to obtain a vector of definitive abstract features, the convolutional neural network trained by comparing a predicted risk factor with a histologically determined Gleason score as ground truth;

identifying a number of pre-stored medical datasets most similar to the current case dataset based on a comparison of the features of the current case dataset with the features of each of the number of pre-stored medical datasets; and

outputting the identified number of most similar pre-stored medical datasets.

2. The method according to claim 1 , wherein obtaining comprises obtaining by the evaluation of the current case and pre-stored medical datasets with respect to tissue abnormalities.

3. The method according to claim 1 , wherein each current case and pre-stored medical dataset comprises multi-parametric MRI data.

4. The method according to claim 1 , wherein each current case and pre-stored medical dataset comprises values of one or more of the following parameters: PSA value, PSA density, DRU score, EPE score, lymph node status, and/or patient age.

5. The method according to claim 1 , wherein each current case and pre-stored medical dataset comprises values of one or more of the following radiologically determined parameters: PI-RADS value, lesion size, lesion location and/or organ volume.

6. The method according to claim 1 , wherein the predefined AI-based method obtains a single scalar value for a risk score as a feature in the evaluation of each current case and pre-stored medical dataset.

7. The method according to claim 1 , wherein the predefined AI-based method obtains, as part of the evaluation, a vector of features comprising values for one or more of the following radiomic parameters: lesion size, lesion intensity, lesion shape, lesion texture, wavelet transformation.

8. The method according to claim 1 , wherein the predefined AI-based method obtains, as part of the evaluation, a vector of features comprising one or more values for a risk score and values for parameters of a task-specific fingerprint.

9. The method according to claim 1 , wherein the most similar pre-stored medical datasets are identified by minimum distance measures between the current case dataset and the pre-stored medical datasets.

10. The method according to claim 2 , wherein obtaining comprises obtaining by the evaluation of the current case and pre-stored medical datasets with respect to tissue abnormalities comprising indications of prostate cancer.

11. An evaluation arrangement for identifying similar pre-stored medical datasets for comparison with a current case dataset, the evaluation arrangement comprising:

a first interface for receiving a current case dataset comprising radiological data of a patient;

a second interface to a number of pre-stored medical datasets each comprising radiological data of another patient;

a processor operating pursuant to instructions stored in a memory, the instruction comprising instruction to:

evaluate each current case and pre-stored medical dataset according to a predefined AI-based method to obtain a number of features for each respective case dataset, wherein the number of features comprise a vector of definitive features including values for one or more of the following radiomic parameters or parameter groups: lesion size, lesion intensity, lesion shape, lesion texture, and wavelet transformation, wherein the predefined AI-based method comprises a convolutional neural network trained by comparing a predicted risk factor with a histologically determined Gleason score as ground truth; and

identify a number of pre-stored medical datasets most similar to the current case dataset based on a comparison of the features of the current case dataset with the features of each pre-stored medical dataset; and

an output interface for outputting the identified number of most similar pre-stored medical datasets.

12. The evaluation arrangement of claim 11 further comprising a screen for displaying the identified datasets.

13. A non-transitory computer-readable medium on which program elements are stored that can be read and executed by a computer, the non-transitory computer-readable medium having stored thereon instructions for:

providing a current case dataset comprising radiological data of a patient;

providing a number of pre-stored medical datasets each comprising radiological data of other patients;

obtaining a number of features including one or more values for a risk score and values for parameters of a task-specific fingerprint for each respective case dataset based on an evaluation of each current case and pre-stored medical dataset according to a predefined AI-based method, wherein the AI-based method comprises a convolutional neural network configured to evaluate datasets to obtain a vector of definitive abstract features, the convolutional neural network trained by comparing a predicted risk factor with a histologically determined Gleason score as ground truth;

identifying a number of pre-stored medical datasets most similar to the current case dataset based on a comparison of the features of the current case dataset with the features of each pre-stored medical dataset; and

outputting the identified number of most similar pre-stored medical datasets.