Treatment planning based on multimodal case similarity
A system, method, and computer program product for treatment planning are disclosed. The system includes at least one processing component, at least one memory component, a training module, a retrieval module, and a plan generator. The training module generates hash codes by hashing features from data sources with data source-specific hash functions, and generates superclass hash codes by hashing the generated hash codes with at least one superclass hash function. The retrieval module extracts features from case data, and locates features from the data sources that are similar to the extracted features. The plan generator calculates outcome probabilities for the case data based on known outcomes associated with the located features.
1. A system, comprising:
at least one processing component;
at least one memory component;
a training module configured to:
generate data source specific hash codes by hashing features from two or more data sources with data source-specific hash functions, wherein there is a different data source-specific hash function for different ones of the two or more data sources; and
generate superclass hash codes by hashing the generated data source-specific hash codes with at least one superclass hash function;
a retrieval module configured to:
extract features from case data; and
locate features from at least one of the two or more data sources that are similar to the extracted features based on a query hash code generated by executing at least one hash function on the extracted features from the case data, and a comparison of the query hash code with one or more of the at least one superclass hash function or the data source-specific hash codes; and
a plan generator configured to:
calculate outcome probabilities for the case data based on known outcomes associated with the located features.
2. The system of claim 1 , wherein the features from the two or more data sources include at least one longitudinal sequence.
3. The system of claim 1 , wherein the extracted features include at least one longitudinal sequence, and wherein the at least one longitudinal sequence is based on a series of medical images.
4. The system of claim 1 , wherein the training module is further configured to generate a machine learning model trained on the superclass hash codes.
5. The system of claim 1 , wherein the retrieval module is further configured to generate the query hash code by hashing a feature from the extracted features with the at least one superclass hash function.
6. The system of claim 1 , wherein the training module is further configured to generate higher level superclass hash codes by hashing a group of the superclass hash codes with at least one higher level superclass hash function.
7. The system of claim 6 , wherein the training module is further configured to generate a machine learning model trained on the at least one higher level superclass hash codes.
8. The system of claim 6 , wherein the retrieval module is further configured to generate the query hash code by hashing a feature from the extracted features with the at least one higher level superclass hash function.
9. A method of treatment planning, comprising:
generating data source specific hash codes by hashing features from two or more data sources with data source-specific hash functions, wherein there is a different data source-specific hash function for different ones of the two or more data sources;
generating hash codes by hashing the generated data source-specific hash codes with at least one superclass hash function;
extracting features from case data;
locating features from at least one of the two or more data sources that are similar to the extracted features based on a query hash code generated by executing at least one hash function on the extracted features from the case data, and a comparison of the query hash code with one or more of the at least one superclass hash function or the data source-specific hash codes; and
calculating outcome probabilities for the case data based on known outcomes associated with the located features.
10. The method of claim 9 , wherein the features from the two or more data sources include at least one longitudinal sequence.
11. The method of claim 9 , wherein the extracted features include at least one longitudinal sequence, and wherein the at least one longitudinal sequence is based on a series of medical images.
12. The method of claim 9 , further comprising generating a machine learning model trained on the superclass hash codes.
13. The method of claim 9 , further comprising generating the query hash code by hashing a feature from the extracted features with the at least one superclass hash function.
14. The method of claim 9 , further comprising generating higher level superclass hash codes by hashing a group of the superclass hash codes with at least one higher level superclass hash function.
15. The method of claim 14 , further comprising generating the query hash code by hashing a feature from the extracted features with the at least one higher level superclass hash function.
16. The method of claim 9 , wherein the superclass hash codes are generated at least by aggregating the data source-specific hash codes by taking a union of the data source-specific hash codes and executing the at least one superclass hash function on the union of the data source-specific hash codes.
17. The method of claim 9 , wherein the at least one superclass hash function comprises a plurality of different superclass hash functions, and wherein each different superclass hash function is associated with a different category of data sources.
18. The method of claim 17 , wherein the different categories of data sources comprise different types of medical imaging modalities such that there is a different superclass hash function for each of the different types of medical imaging modalities.
19. A computer program product for treatment planning, the computer program product comprising a computer readable storage medium having program instructions embodied therewith, the program instructions executable by a processor to cause a device to perform a method, the method comprising:
generating data source specific hash codes by hashing features from two or more data sources with data source-specific hash functions, wherein there is a different data source-specific hash function for different ones of the two or more data sources;
generating hash codes by hashing the generated data source-specific hash codes with at least one superclass hash function;
extracting features from case data;
locating features from at least one of the two or more data sources that are similar to the extracted features based on a query hash code generated by executing at least one hash function on the extracted features from the case data, and a comparison of the query hash code with one or more of the at least one superclass hash function or the data source-specific hash codes; and
calculating outcome probabilities for the case data based on known outcomes associated with the located features.
20. The computer program product of claim 19 , further comprising generating higher level superclass hash codes by hashing a group of the superclass hash codes with at least one higher level superclass hash function.