Diagnostic system and diagnostic method
A first interface for reading a medical patient image record is provided. Furthermore, provision is made of an encoding module for machine-based learning of data encodings of image patterns by an unsupervised deep learning and for establishing a deep-learning-reduced data encoding of a patient image pattern contained in the patient image record. Furthermore, provision is made of a comparison module for comparing the established data encoding with reference encodings of reference image patterns stored in a database and for selecting a reference image pattern with a reference encoding which is similar to the established data encoding. An assignment module serves to establish a key term assigned to the selected reference image pattern and to assign the established key term to the patient image pattern. A second interface is provided for outputting the established key term with assignment to the patient image pattern.
1. A computer-assisted diagnostic system for assisting a medical diagnosis, comprising:
a) a patient image record reader, wherein said patient image record reader is used for reading a patient image record,
b) a machine-based learning encoding module, wherein said machine-based learning encoding module is used for data encodings of image patterns by unsupervised deep learning using unclassified training image patterns, and for establishing a deep-learning-reduced data encoding of a patient image pattern contained in the patient image record,
c) a medical reference image pattern database, wherein said database for storing a multiplicity of medical reference image patterns with assigned key terms relevant to the medical diagnosis,
d) an image pattern comparison module, wherein said comparison module is used for comparing the established data encoding with reference encodings of reference image patterns stored in the database on the basis of a similarity measure and for selecting a reference image pattern from the medical reference image patterns with a reference encoding which is similar to the established data encoding in respect of the similarity measure,
e) a key term assignment module, wherein said assignment module is for establishing a key term assigned to the selected reference image pattern and for assigning the established key term to the patient image pattern, and
f) a key term output, wherein said key term output is assigned to the patient image pattern.
2. The diagnostic system as claimed in claim 1 , wherein the machine-based learning encoding module comprises a neural autoencoder for establishing the reduced data encoding.
3. The diagnostic system as claimed in claim 2 , wherein the neural autoencoder comprises an input layer for entering raw image data, low-level image features and/or image features learned by a convolutional neural network.
4. The diagnostic system as claimed in claim 1 , wherein the multiplicity of medical reference image patterns comprises a training set for deep-learning-based training of the encoding module.
5. The diagnostic system as claimed in claim 1 , wherein the key terms relevant to the diagnosis specify local properties of an associated reference image pattern and/or image pattern-overarching information about a patient.
6. The diagnostic system as claimed in claim 1 , wherein the similarity measure is a machine-learned similarity measure.
7. The diagnostic system as claimed in claim 1 , further comprising a segmentation module for establishing a position, orientation and/or dimension of organs, organ parts and/or organ regions and for segmenting the patient image record into patient image segments specific to organs, organ parts and/or organ regions.
8. The diagnostic system as claimed in claim 7 , wherein the machine-based learning encoding module is configured to establish segment-individual data encodings of patient image segments.
9. The diagnostic system as claimed in claim 7 wherein the reference image patterns with assigned key terms relevant to the diagnosis comprises a multiplicity of reference image segments with assigned segment-specific key terms.
10. The diagnostic system as claimed in claim 1 , wherein the assignment module is configured to assign a relevance value to the established key term in a manner dependent on the similarity measure between the established data encoding and the reference encoding similar thereto and/or in a manner dependent on a frequency with which this key term is established.
11. The diagnostic system as claimed in claim 10 , wherein an aggregation module (AGM) for aggregating key terms established for different patient image segments to form an aggregated key term (KA) in a manner dependent on the relevance values of the established key terms.
12. A computer-assisted diagnostic method for assisting a medical diagnosis, wherein
a) a patient image record is read,
b) data encodings of image patterns are learned by machine by means of unsupervised deep learning using unclassified training image patterns, and a deep-learning-reduced data encoding of a patient image pattern contained in the patient image record is established,
c) the established data encoding is compared to reference encodings of stored medical reference image patterns on the basis of a similarity measure, wherein key terms relevant to the diagnosis are assigned to the medical reference image patterns,
d) a reference image pattern from the medical reference image patterns with a reference encoding similar to the established data encoding in respect of the similarity measure is selected,
e) a key term assigned to the selected reference image pattern is established and assigned to the patient image pattern, and
f) the established key term is output with assignment to the patient image pattern.
13. A computer program product, comprising a non-transitory computer readable hardware storage device having computer readable program code stored therein, said program code configured to execute a diagnostic method as claimed in claim 12 or to implement a diagnostic system.
14. A computer-readable storage medium, comprising a stored computer program product stored on a non-transitory computer storage medium as claimed in claim 13 .