Medical ontologies for machine learning and decision support

• US 7,899,764 B2
• Filed: 06/15/2007
• Issued: 03/01/2011
• Est. Priority Date: 02/16/2007
• Status: Active Grant

• Current Assignee: Siemens Healthineers AG (Siemens AG)
• Original Assignee: Siemens AG
• Inventors: Zillner, Sonja, Tsymbal, Alexey, Martin, Huber
• Primary Examiner(s): HOLMES, MICHAEL B
• Application Number:
  US11/763,717
• Publication Number:
  US 20080201280A1
• Time in Patent Office: 1,355 Days
• Field of Search: 706/12
• US Class Current: 706/12
• CPC Class Codes: G06N 20/00 Machine learning
• CPC Class Codes: G16H 10/60 for patient-specific data, e.g. for electronic patient records
• CPC Class Codes: G16H 30/20 for handling medical images, e.g. DICOM, HL7 or PACS
• CPC Class Codes: G16H 50/20 for computer-aided diagnosis, e.g. based on medical expert systems
• CPC Class Codes: G16H 50/70 for mining of medical data, e.g. analysing previous cases of other patients
• CPC Class Codes: G16Z 99/00 Subject matter not provided for in other main groups of this subclass

Abstract

A medical ontology may be used for computer assisted clinical decision support. Multi-level and/or semantically grouped medical ontology is incorporated into a machine learning algorithm. The resulting machine-learnt algorithm outputs information to assist in clinical decisions. For example, a patient record is input to the algorithm. Based on the incorporated medical ontology, similarities are aggregated in different groups. An aggregate similarity of at least one group is a function of an aggregate similarity of at least another group. One or more similar patients and/or outcomes are identified based on similarity. Probability based outputs may be provided.

Claims

1. A method for use of a medical ontology for computer assisted clinical decision support, the method comprising:

providing a medical ontology indicating one or more semantic groupings of features;

learning, by a processor, to identify at least one similar patient record from a set of patient records, the learning being a function of the one or more semantic groupings of features of the medical ontology; and

outputting information derived from the learning.

2. The method of claim 1 wherein learning comprises integrating a graph structure of the medical ontology, the one or more semantic groupings of features corresponding to one or more respective branches of the graph structure.

3. The method of claim 2 wherein the features represent heterogeneous concepts, and wherein a first branch comprises a first type of data selected from the group of:

clinical, imaging, ECG, and genetics, and a second branch comprises a second type of data selected from the group.

4. The method of claim 1 wherein learning as a function of the one or more semantic groupings comprises constraining a contribution of the features to a distance with the one or more semantic groupings.

5. The method of claim 4 wherein constraining comprises aggregating features into a grouping contribution within each semantic grouping, the distance being a function of the grouping contribution.

6. The method of claim 1 wherein learning as a function of the one or more semantic groupings comprises weighting a contribution of each semantic grouping relative to the other semantic groupings.

7. The method of claim 6 further comprising:

receiving user input of the relative weighting;

wherein weighting comprises weighting as a function of the user input of the relative weighting.

8. The method of claim 6 further comprising:

optimizing contributions such that the at least one similar patient record is more similar.

9. The method of claim 1 wherein learning as a function of the one or more semantic groupings comprises learning from a multilevel grouping structure of the semantic groupings.

10. The method of claim 9 wherein each semantic grouping for a given level of the multilevel grouping structure has a relative contribution for the given level, the relative contribution constraining the learning.

11. The method of claim 1 wherein providing the medical ontology comprises combining a plurality of types of ontologies into the medical ontology, at least one of the one or more semantic groupings corresponding to different ones of the types of ontologies.

12. The method of claim 1 wherein learning comprises creating a machine-learnt algorithm, and wherein outputting comprises:

applying the machine-learnt algorithm to a patient record;

identifying the at least one similar patient record with the machine-learnt algorithm; and

outputting an identification or data for the at least one similar patient record.

13. The method of claim 1 wherein outputting comprises outputting validation results.

14. The method of claim 1 wherein the medical ontology comprises a range for at least one feature;

further comprising;

removing data with a value for the at least one feature outside of the range before learning or during application of a machine learned algorithm derived from the learning.

15. The method of claim 1 wherein the medical ontology comprises an identification of correlated or redundant first and second features;

further comprising;

reducing a weighting and/or removing feature consideration of a first feature relative to a second feature.

16. A system for use of a medical ontology for computer assisted clinical decision support, the system comprising:

a memory operable to store a machine-learnt algorithm, the machine-learnt algorithm integrating a multi-level medical ontology, the multi-level medical ontology having a hierarchal node structure defining relative contribution of features at different levels of the multi-level medical ontology; and

a processor operable to apply the machine-learnt algorithm to a medical record of a patient.

17. The system of claim 16 wherein the machine-learnt algorithm is operable to identify a previous medical record similar to the medical record of the patient.

18. The system of claim 16 wherein the multi-level medical ontology includes features with common semantics for a first type of node and second types of nodes connected with at least a first type of node, and wherein the processor is operable to aggregate a contribution of the features for each of the first types of nodes, and aggregate a contribution of the contributions of the first types of nodes for each of the second types of nodes.

19. The system of claim 16 further comprising:

a display operable to display a representation of the multi-level medical ontology, the display including weights for contributions of features and/or nodes.

20. In a computer readable storage media having stored therein data representing instructions executable by a programmed processor for computer assisted clinical decision support with a medical ontology using machine learning, the storage media comprising instructions for:

incorporating, into a machine learning algorithm, domain knowledge from a medical ontology describing interrelations between features;

searching, with the machine learning algorithm, a sequence of instances where each instance is represented by at least one single feature vector describing the instance, the searching including aggregating similarities in different groups, an aggregate similarity of at least one group being a function of an aggregate similarity of at least another group; and

identifying, in response to the searching, a similar instance with the machine learning algorithm.

21. The instructions of claim 20 wherein incorporating comprises defining a distance function of the machine learning algorithm by a distribution of weights as a function of the medical ontology.

22. The instructions of claim 20 further comprising:

receiving user input adjusting the medical ontology; and

re-performing the incorporation as a function of the adjusted medical ontology.